BeeKeep3r: feature catalogue

Every function of the IDE and the Harness.

168 features of the BeeKeep3r IDE and 153 features of the BeeKeep3r Harness, grouped by area. Each feature lists its individual functions.

BeeKeep3r IDE

168 features · 22 areas

To the IDE

Start, projects and navigation

Getting started with BeeKeep3r IDE: open and clone projects, run several projects in parallel and jump quickly between projects, Worktrees and sessions.

  • Launcher / start screen

    Start screen 'Build the future.' with an orbit of the supported coding agents and quick-access shortcuts. Takes new users straight to an open project.

    Functions New Workspace · Agent orbit · Open Folder · Chat · Agent Swarm · Clone Repo · Multi Project

  • Clone repository

    Clones a Git repository and opens it immediately as a workspace. Saves the detour via the terminal.

    Functions GitHub repositories · Git URL · Folder name · Error help

  • Open project / New project

    One dialog for opening a new project (new Git repo), a local folder, a Git URL or a GitHub repo. Also used when creating new agents and chats.

    Functions New project · Local folder · Git URL / GitHub repository · Keyboard operation

  • Recent projects (open several at once)

    List of recently opened projects with checkboxes for opening several at once.

    Functions Multi-select · Add folders… · Broken paths

  • Multi-Project Workspace (workspace bar)

    Several projects open in parallel, each with its own isolated agent sessions. Project chips show status and waiting sessions.

    Functions Project chips · Show/hide and Solo · Focus / Pin · New menu · Soft-cap notice · Close with confirmation

  • Project layouts: Focus, Columns, Rows, Grid

    Arranges all open projects as tabs (Focus), columns, rows or tiles and switches all panes between agent and terminal view.

    Functions Focus · Columns / Rows · Grid · Agent/Terminal view · Same layout for agent and terminal

  • Project colours

    Each project is given a colour that appears in the pill, tab dot and pane border, so that parallel projects remain distinguishable.

    Functions 12 colour presets · Custom colour

  • Project tree in the sidebar

    A tree of project › Worktree › session/chat for agent and terminal view, searchable and groupable.

    Functions Grouping · Search · Create Worktree · Pin / Settled · Hide / Solo · Markers · Rename chats

  • Jump field / Command Bar (⌘K)

    One input field for jumping to projects, Worktrees, sessions and commands and for running all quick actions. Replaces the former command palette.

    Functions Jump to · Quick actions · Layout actions · Shortcut reference · Browse commands… · Free text as a command · Target picker

  • New session (session launcher)

    Starts one or more agent/terminal sessions in a project, optionally in an isolated Git Worktree, with a preset and an optional task.

    Functions Presets · Git mode · Isolated Worktree · Branch collision · Task · Swarm fields

  • Session overview

    Table of all sessions across all open projects with status, branch, context and last activity.

    Functions Columns · Waiting sessions · Jump directly

  • Restoring open projects

    On restart, BeeKeep3r IDE reopens the most recently open workspaces with their layout; terminals only start after successful sign-in and licence check.

    Functions Layout persistence · Start gate

  • Dashboard

    Overview of workspaces, terminals, installed agents and account usage.

    Functions Runs this month · Installed agents · Live terminals · Workspaces · Session activity

  • Status bar (footer)

    Shows terminals, agents, provider quotas, estimated token costs, memory usage and the active GitHub account.

    Functions Usage · Tokens & cost · Resources / Memory · Counters · GitHub accounts

BeeSpace, terminals and editor

The work area: panes and splits, terminals, browser, editor and the right-hand side panel.

  • BeeSpace: panes and splits

    Split work area in which each pane can be a terminal, an agent, a file explorer, a browser or an agent thread.

    Functions Pane types · Local agents · Split · Tabs · Switch panes · Solo · Maximize · Close with confirmation

  • Starting agents and terminals ('What do you want to run?')

    In an empty project, select and start several agents and terminals at once, including your own CLI commands.

    Functions Multi-select · Custom command · Set up a swarm…

  • Terminal pane menu

    All actions relating to a terminal pane: rename, restart the agent, export the session, layout and view.

    Functions Rename · Restart agent / Stop agent · Change working folder · Export session as Markdown · Copy / Paste / Clear · Icon & colour… · View · Checkout / Worktree · Solo – only this project · Context indicator · Status indicator

  • Layout presets, freeform canvas, Stage and minimap

    Arrange panes freely on a zoomable surface or tidy them up with a preset; Stage shows a full-screen arrangement.

    Functions Organize presets · Stage · Zoom and pan · Minimap · Esc chain

  • Browser pane (freeform browser node)

    A real embedded browser (native web view) as a pane in the BeeSpace, e.g. for the local dev server.

    Functions Navigation · Zoom · Viewport presets · Dev server detection

  • Right panel (Git, Editor, Commands, Browser, Skills, Scripts, Lens, Context)

    A dockable side panel with tools for the focused pane.

    Functions Docking · Maximise / size · Commit, push & PR · Browser panel

  • Editor with file explorer

    Code editor (CodeMirror) with syntax highlighting, a resizable file tree, tabs and safe saving.

    Functions File tree · Tabs · Autosave · Languages

  • Terminal settings

    Font, input, copy behaviour, links and accessibility of the terminals, plus editor autosave.

    Functions Font family · Font size with reset · Shift+Enter inserts a line break · Copy on select · Copy text, not the frame · Programs may write to the clipboard · Save editor changes automatically · Open links in the in-app browser · Screen reader mode · Typography · Shift+Enter · OSC 52 · Links in the in-app browser · Editor autosave

  • Ghost completion (autocomplete)

    Inline suggestions in the terminal from the local command history, grouped by working directory; the history stays on your machine.

    Functions Show ghost text · Record command history locally · Statistics · History by directory · Clear history · Ghost text · Local history

Agents and chat

Structured agent chats with coding CLIs: agents as folders for chats, a composer with attachments and references, and control over approvals, changes and context.

  • Agent view: agents and chats

    Agents are named folders per project in which chats live. Each agent has an icon, a colour and Memory, Skills and Settings pages.

    Functions Create agents · Icon and colour · Manage chats · Agent pages · Search · Multi-project layout

  • Agent composer (input)

    Input field for agent chats with send/stop, history, and model, effort and permission selection.

    Functions Send / Stop · Input history · Model selection · Effort · Permission mode · Workspace selection · Shortcuts in the field

  • Attachments in chat

    Attach images, PDFs, Markdown, text and code to a message.

    Functions Drag & drop, click, paste · File types · Limit

  • @ references to files

    Use @ to search for a file or folder in the project and insert it as a reference that the CLI understands.

    Functions File and folder search

  • @plugin in agent chat

    Enable connected plugins (e.g. GitHub, Linear, Notion) via @ for a single message, so that the agent may use the connected account.

    Functions Plugin reference

  • Context indicator and Compact

    Shows how full the context window is and summarises the conversation so far, on request or automatically.

    Functions Context ring · Manual Compact · Auto-Compact (opt-in) · View summary

  • / commands and skills in chat

    Insert commands and skills from the project and the user, as well as built-in commands, via '/'.

    Functions Sources · Add to Studio · Built-in commands · Prerequisite

  • Message queue and Steer

    While the agent is working, queue further messages or feed them directly into the running turn.

    Functions Queue · Steer · Manage queue

  • Approvals, tool activity and sub-agents

    Transparency about everything the agent does: grouped tool steps, approval requests, sub-agents and background tasks.

    Functions Approval card · Tool groups · Sub-agents · Background tasks · Run indicator · Copy message

  • Changes per turn, diff and rollback (checkpoints)

    Shows which files an agent turn has changed, with a diff per file and the option to roll the turn back.

    Functions File tree of changes · Diff per file · Rollback

  • Response rendering: Markdown, diagrams, HTML preview

    Formatted responses with copyable code, tables, Mermaid diagrams and a safe preview of agent-generated HTML pages.

    Functions Code and tables · Mermaid · Images · HTML preview

  • Context panel of the chat

    Everything about the chat's environment at a glance: Worktree, branch, changes, pull requests, connectors, context, outputs and commands.

    Functions Environment · Changes · Pull requests · Connectors · Context / Outputs · Commands & Skills

  • Terminal in chat and project scripts

    A terminal in the chat's checkout below the conversation, plus your own project scripts as one-click buttons.

    Functions Terminal drawer · Project scripts · Setup scripts

  • Open in (external apps)

    Opens the chat's checkout in an installed editor, file manager, terminal or Git client.

    Functions App categories · Copy path

  • Chat ↔ terminal handoff and transcript view

    Switch a Claude Code session seamlessly between chat view and terminal (claude --resume); a terminal Claude can be viewed as a readable transcript.

    Functions Continue in terminal · Continue as chat / Take back · Transcript view

  • Chat mode ('Just chat')

    A plain conversation with a model of your choice, without a project, file access, commands or tools.

    Functions Separate chat list · Scope notice

  • Supported agent programs (coding CLIs)

    BeeKeep3r IDE detects installed coding CLIs and starts them as a terminal agent or as a structured chat.

    Functions Terminal agents · Structured agent chat · Local models · Agent hooks

Swarms

Several agents with roles (Coordinator, Builder, Scout, Reviewer) work in parallel on a shared mission.

  • Setting up a Swarm (Mission, Roster, Launch)

    In three steps, describe a mission, put together a team of agents and start the Swarm. Suited to larger tasks that can be parallelised.

    Functions Roles · Quick presets · Agent per seat · Auto-approve · Launch mode · Swarm skills · Context files

  • Swarm cockpit (running Swarm)

    Live overview of a Swarm as an agent graph with plan, chat, activity and roster, plus control via direct message.

    Functions Lifecycle · Time budget · Tabs · Quick commands · Per agent · Attach swarm · Interruption report

Routines

Scheduled, non-interactive commands (e.g. agent dispatch) that BeeKeep3r IDE runs in a folder at fixed intervals.

  • Routines (scheduled agent work)

    A routine runs a shell command in a folder every N minutes and logs the result and output. Typical use: automatically distributing tickets labelled 'agent-ready' to agents.

    Functions Interval schedule · Run now · Pause / Resume · Result log · Safeguards · Board dispatch preset

Git integration

Git directly in BeeKeep3r IDE: create a repository, Worktrees for parallel agents, commit/push/PR, review and merge pull requests.

  • Initialising Git and first commit

    Turns a plain folder into a Git repository and guides you through the first commit, which unlocks branches and Worktrees.

    Functions git init with binding · First commit

  • Worktrees and branches per chat

    Each chat can work in its own Worktree or branch, so that parallel agents do not get in each other's way.

    Functions Worktree modes · Fetch first · Sync status · Local and remote branches · Checkout bar · Clean-up

  • Git panel

    Complete Git interface: stage, commit and push changes, branches, stash and conflict resolution.

    Functions Changes / History · Push / Pull · Branches · Stash · Conflicts · Protect default branch · GitHub search

  • Commit, push & PR (context-dependent Git button)

    One button whose action depends on the state of the repo: Commit, Commit & push, Commit, push & PR, Pull or View PR.

    Functions State-dependent action · Safety prompt · Locks with stated reasons

  • Pull requests: check out and merge

    See open PRs with check status, check out a PR locally and merge it directly from BeeKeep3r IDE.

    Functions PR list · Check out PR · Merge dialog · GitHub account per project

  • Worktree checkout banner

    Indicates that a Worktree is currently being shown instead of the full project, and switches to the main checkout.

    Functions Badges · Switch

Scripts and security

Start project scripts with one click, with risk levels, and protection against catastrophic agent commands.

  • Scripts Quick Access with risk levels

    Lists all executable scripts of a project (package.json, shell, PowerShell, Python, your own project scripts) and starts them in a separate terminal. A static analysis classifies each script as Normal, Caution or Destructive.

    Functions Detection without execution · Risk level Normal · Risk level Caution · Risk level Destructive · Details · Status · System check · Rescan

  • Command Guard

    A second layer of security that blocks catastrophic, irreversible shell commands from Claude Code and Codex agents before they run.

    Functions Blocked patterns · Shell-aware · Fail-open

Voice

Speech as input and output: dictation, voice commands, Full voice, Voice Mode with Spheres and the associated settings.

  • Voice modes: Dictate, Command, Full voice

    Three modes, each with its own key: Dictate writes into the focused field or terminal, Command controls the app, and Full voice is a live conversation with the assistant, which can do everything Command can.

    Functions Dictate · Command · Full voice / Live AI · Voice keys · Microphone menu · Microphone selection · Shortcut

  • Voice commands, recipes and voice tools

    Fixed voice commands respond immediately; compound sentences ('new workspace named Bee One with Claude') are recognised as recipes; everything else goes to a voice assistant with tools.

    Functions Fixed commands · ⌘K by voice · Recipes · The assistant's voice tools · Safety rules · Who answers

  • JEV app control (AI-assisted voice control)

    Understands natural spoken instructions and selects suitable app actions and targets; JEV never supplies code or text to be typed.

    Functions Action selection · Privacy notice · On-device recognition

  • Voice Mode in agent chat (Orb / Spheres)

    Talk to the agent instead of typing: an animated orb shows listening, working and speaking, replies are read aloud and tasks are handed over to the agent.

    Functions Orb designs (Spheres) · Hands-free / Push-to-talk · Live conversation · Read aloud / show text · Agent steps · Replies in the background · Queue

  • Dictation in the chat composer

    Dictate a message into the composer, with a level meter and a choice between inserting, sending or discarding.

    Functions Dictation bar

  • Voice settings

    Brings together voice input, voice control, live conversation and voice output; each mode has its own key with hold or latch behaviour.

    Functions Enable voice · Input device · Voice keys · Live conversation engine · Speak replies & callouts · Reply style · Local audio models · On-device speech recognition · Voice output · Cloud · Languages · Style · Sounds and progress

Music and media

Music and videos alongside your work: radio, YouTube, Spotify and the DANIC channel, with a mini player and floating video.

  • Music pane (Station, YouTube, Spotify, Apple Music)

    A music pane in the BeeSpace with four source tabs; playback continues when you switch workspace and layout.

    Functions Station · YouTube · Spotify · Apple Music · Transport

  • Floating video and mini player

    YouTube videos keep playing without reloading: docked in the Music pane, floating freely or as a mini player in the title bar.

    Functions Docked / Floating · Mini player · Dedicated transport bar

  • YouTube favourites and DANIC channel

    Save videos and playlists locally as favourites and play the DANIC YouTube channel with one click.

    Functions Favourites · DANIC on YouTube

Providers and models

Central place for subscriptions, your own API keys (BYOK), local models and the assignment of models to chat, terminals and voice.

  • Providers › Subscriptions

    Connects existing coding-agent subscriptions via their official apps: Claude Pro/Max, ChatGPT Plus/Pro (via the Codex CLI) and the GLM Coding Plan from Z.AI.

    Functions Claude Pro / Max · ChatGPT Plus / Pro · GLM Coding Plan · Status indicator

  • Providers › API connections (your own API keys)

    Connect your own API accounts or gateways; 26 templates plus any OpenAI-compatible endpoint. Keys are encrypted locally and never returned to the interface.

    Functions Provider catalogue · Edit connection · Manage keys · Saved connections · Found on this device · JEV (Voice control)

  • Model routing

    Defines which model the BeeKeep3r IDE chat uses, which local model coding agents run on and where terminals and voice route to.

    Functions BeeKeep3r IDE chat · Coding agents on a local model · Local terminals and voice · Harness

  • Local models (Ollama / LM Studio)

    Bind models that run on your own machine to chat, terminals and voice; Ollama and LM Studio are connected via their OpenAI-compatible server.

    Functions Runtime endpoint · Bind models · Bind by path · ASSIGNMENT

  • Models and agents (Agent hooks, resources, limits, Command Guard)

    Hooks for finished/waiting signals from all CLIs, resource switches, a time budget for Swarms and the Command Guard against catastrophic shell commands.

    Functions Notify when agents finish (Agent hooks) · Codex notes · Clean up backups · Memory tools for agents · Clean up finished agents · Time budget / Settle idle swarms automatically · Command Guard · Agent hooks · Swarm limits

  • Claude connection (for Harness terminals)

    Defines what Claude is signed in with in Harness terminals: a Claude subscription via setup-token or an Anthropic API key. Without this connection, Harness cannot start Claude.

    Functions Subscription connection · API key · Status · Remove connection

  • MCP servers

    One MCP configuration (~/.beemind/mcp.json) as the source, which is distributed to all detected coding agents.

    Functions Source · Sync · Tool detection

  • Skill sync

    Gives the agents the BeeKeep3r IDE skills or brings their skills back, and offers a reset of your own skills, agents and commands.

    Functions Sync per skill · Remove from all tools · Your own skills, agents and commands (Reset)

  • Connect your coding agent (setup assistant)

    Guided setup of Claude Code, Codex or Cursor with an installation command to copy and a detection check.

    Functions Claude · Codex · Cursor · Installation guide

Plugins, skills and Studio

Extend BeeKeep3r IDE: plugins, your own skills, agents and commands, and the Command Builder.

  • Plugins (connector catalogue)

    Catalogue of around 50 services that can be connected via API key or OAuth; credentials stay in the keychain.

    Functions Catalogue · Connection status · Secure storage

  • Harness Computer Use

    Agent-neutral MCP runtime for desktop control (Claude, Cursor, Codex, OpenCode, Cline): semantic snapshots, actions restricted to them and redaction of protected values.

    Functions Status indicator · Platform

  • Skills

    List of the bundled Harness skills and your own skills, with search; skills can be dragged onto a terminal to insert them.

    Functions Categories · Locked entries · Reorder

  • New skill (assistant), editing and deleting a skill

    Describe your intent, an installed agent CLI drafts the SKILL.md, you review and create it; optional sync to the detected agent tools.

    Functions AI draft · Sync to agent tools · Edit · Delete

  • Studio (Skills, Agents, Commands)

    One list page per type with the tabs Mine, BeeKeep3r Harness and All, cards with a best-practice score and a sidebar for Setup Health, sync targets and topics.

    Functions Cards · Duplicate to mine · SETUP HEALTH · SYNC TARGETS · Topics · Side panel New menu

  • Best-practice engine and pattern catalogue

    A lint engine checks agents, skills and commands against published best practices from Anthropic and OpenAI (rules BP-01 to BP-08) and proposes corrections as a diff; ten patterns provide proven structures.

    Functions Rules · Auto-fixes · Patterns Agents · Patterns Skills · Patterns Commands · Source references

  • New agent (assistant)

    Four steps from the described purpose to the finished subagent, with pattern, live best-practice check and sync.

    Functions Pattern suggestion · Live checks · Sync targets

  • Import setup

    Bring in an existing agentic setup from local folders or a Git repository, with security check, sorting by topic, best-practice check and sync.

    Functions Sources · Security check SC-01 to SC-08 · Topics on import · Apply best-practice fixes automatically · Result

  • Visualise

    Shows the flow of a command as an animated flow graph (looping simulation), even if it is only derived from the text.

    Functions Derived flow · Make explicit · Error hints

  • Command Builder and Command Editor

    Visual editor for your own slash commands: drag building blocks onto a canvas, connect them, define gates, check plausibility live, simulate, save, sync and run.

    Functions Building blocks · Gate assistant · Insights collector · Model per step · Skills per step · Plausibility lint PL-01 ff. · Org policy · Inspector with Markdown · Simulate / Run · Canvas controls

  • Studio Reset

    Removes all skills, commands and agents that are your own, imported or created in Studio, including synchronised copies, with a backup.

    Functions Scope · Protection

Remote, Co-op and Mobile

Agents on remote machines, shared sessions in the team and control from your smartphone.

  • Remote hosts

    Set up your own servers via SSH as a target for agents. BeeKeep3r IDE uses the system ssh, the ssh-agent and ~/.ssh/config; it does not store a password and does not read a private key.

    Functions Host assistant (3 steps) · Connection check · Host key verification · Prepare host · Sessions · Edit / Delete / Test connection

  • Starting an agent on a remote host

    When starting an agent, you choose Local or a remote host; terminal and chat then run on the server and are mirrored live.

    Functions Remember start target · Remote badge · Cockpit notice and pull · Remote chat

  • Co-op

    Shares a live view of your own work (projects, terminals, chats) with an org member as View or Interact, end-to-end encrypted and secured by a security code.

    Functions Roles · Sharing dialog · Security code · Host banner · Input requests · Reduce shared items afterwards · Guest view · Requests and invites (Inbox) · Presence cursors · This device

  • BeeFast (Mobile)

    Connects the computer to the BeeFast app via an encrypted relay; the phone is paired by one-time code or QR code.

    Functions Connection · One-time code and QR · Paired devices

  • What the smartphone can do

    The phone sees the desktop's projects, sessions, agents and activity, and can open and create chats and terminals, send prompts, type in the terminal and start Harness commands.

    Functions Snapshot · Create sessions · Send prompts · Terminal live · Harness commands · Disconnect device

Memory, Board and notifications

Notes, tasks and all of the app's notices in one place.

  • Memory (notes hub with graph)

    A 'second brain' in the file system: Markdown notes in the .beememory folder with wikilinks, sections, backlinks and a graph. Agents read and write the same notes via a bundled MCP server.

    Functions Notes · Wikilinks · Sections · Inspector · Graph · Multiple hubs · Agent access · Obsidian-compatible

  • Notification Center

    Collects messages from chats and terminal agents; one click jumps to the session concerned.

    Functions Kinds · Unread · Focus rule · Navigation

  • Notifications

    Controls the sound, the attention sound, system notifications and which kinds of message appear; the Notification Center history can be cleared.

    Functions Play a sound · Sound when an agent waits for you · Show system notifications · Notifications for chats and terminal agents · KINDS · SWARM ROLES · Clear history · Sound · Waiting-for-you sound · System notifications

Onboarding, feedback and updates

Guided tours, feedback to DANIC, diagnostics and automatic updates.

  • BeeKeep3r IDE product tour

    A one-minute tour of the IDE, narrated to match the CLIs on the machine (Claude workflow, Cursor workflow or starting without a CLI).

    Functions Stops · Variants · Persistence

  • Onboarding hub (Guided tours)

    Overview of all tours with phases, duration and progress; workflow tours simulate complete Harness runs, interface tours highlight the real IDE.

    Functions Workflow tours · Interface tours · Progress · Recap

  • Feedback bubble (Report a bug / Feature request)

    Report bugs or ideas directly from BeeKeep3r IDE, optionally with images and a stack trace; reports go to DANIC and are visible to org owners in BeeKeep3r SaaS under Feedback & Ideas.

    Functions Two steps · Images · Diagnostics · Confirmation · Draft · Hide bubble

  • Diagnostics and support contact

    Open logs, switch on detailed logging, generate a diagnostics bundle without tokens and contact support by email.

    Functions Log file / Open logs · Detailed logging · Diagnostics bundle · Contact support

  • App updates (auto-update)

    BeeKeep3r IDE checks for new versions (optionally on every launch), downloads them with a progress indicator, shows release notes and installs after a restart.

    Functions Check for updates on launch · Update channel · Update dialog · Installation that cannot update itself · Last check

  • About and Third-party licenses

    List of all included open-source components with licence text, searchable and filterable.

    Functions Search and filter · Licence text

Harness in the IDE

How BeeKeep3r IDE shows and starts the BeeFlows of the BeeKeep3r Harness: product pages, CommandBar, target terminal, Workflow guide.

  • Harness Overview

    Product page for Harness with the core message, catalogue figures, a layer diagram and the benefits. For free users it is the entry point for unlocking; for Harness customers it is a status display.

    Functions Hero with claim · Harness catalogue (figures) · System layers · Why Harness · Unlock/Manage CTA · Industry alias chip

  • Commands (Bees & BeeFlows catalogue)

    The curated catalogue of all Harness commands: two delivery workflows (Greenfield, Brownfield) and 16 command clusters, each with a 'use when' explanation. Commands can be searched, viewed as a flow and inserted directly into a terminal.

    Functions Delivery workflows · Command clusters · Search · Row status · Entitlement summary · More · Command canvas · Flow graph

  • Agents (Bees)

    Explains which main agents work behind a request and how Harness routes a request via the ops-orchestrator into Foundation, Legacy or Delivery work.

    Functions Request routing · Main agents

  • System Visualisation (BeeOffice)

    Embeds the interactive system visualisation of the BeeKeep3r Harness (BeeOffice) in BeeKeep3r IDE as a native webview window.

    Functions Embedded view · Offline/error state · Launcher mode

  • Feedback & Help (Harness portal)

    Shows the feedback/ideas page and the help from the customer portal directly in BeeKeep3r IDE. Your own feedback is still submitted via the feedback bubble; here you see what you have submitted.

    Functions Feedback · Help · Feedback portal · Help portal

  • CommandBar: Commands pill and command search

    In the ⌘K menu, Harness commands can be browsed by category or searched globally and transferred to the chat/terminal input with Enter.

    Functions Category pills · Two-column view · Global search · Keyboard control

  • Choose target terminal + launch stepper

    After a command is selected, BeeKeep3r IDE asks which open terminal it should go to, or starts a new session in three steps into which the command is inserted.

    Functions Open terminals · New terminal… (Stepper) · Safe to cancel · Error handling

  • Commands in the Right Panel + footer button

    The command catalogue as a side panel next to your work: the same 16 categories, one click copies the command to the clipboard.

    Functions Copy instead of Insert · Same count · Footer trigger

  • Slash menu '/' and 'Harness commands' in the window menu

    When you type '/', BeeKeep3r IDE lists the Harness commands installed in the workspace; the window menu of a terminal has a filterable command picker.

    Functions Slash menu · OS command picker

  • Workflow guide (OS guide bar)

    A small bar with the recommended command order for four situations; each step is a chip that inserts its command.

    Functions New product · Feature · Bug · Hotfix · Switch off permanently

  • Terminal flow companion (Show BeeFlow)

    Shows the flow graph of the most recently inserted Harness command below a terminal, simulated in step with the terminal activity.

    Functions Flow of the command · Flow selection · Size · Simulation · Enlarge / Shrink / Close

Harness connection, Runtime and projects

Connecting with the DANIC ID, BeeKeep3r Runtime, OS updates and linking a project to Harness.

  • Connection (connect with DANIC ID)

    Connects the signed-in DANIC ID to BeeKeep3r Harness: the licence is fetched and stored in the OS keychain, and the local BeeKeep3r Runtime delivers the unlocked commands.

    Functions Connect with DANIC ID · Status display · Reconnect · Continue onboarding / Start for free / Upgrade · Disconnect · Connect / Disconnect · Start for free / Upgrade / Continue onboarding · Onboarding browser · Error messages

  • Runtime (BeeKeep3r Runtime, OS pack, Runtime check, diagnostics)

    Shows the state of the local BeeKeep3r Runtime per workspace (OS version, source, channel, pack cache), checks in five steps whether commands reach a Claude terminal, and repairs the wiring.

    Functions BeeKeep3r Runtime (status) · BeeKeep3r Harness channel · Runtime check · Rewrite commands / Connect · Diagnostics · Runtime status · Rewrite commands

  • BeeKeep3r Harness updates (automatic / manual)

    Users decide whether new BeeKeep3r Harness versions apply automatically to new terminals or whether the current version is kept and an update is activated deliberately.

    Functions Automatic · Manual · Version display · Activate · Errors & fallback · Minimum security version · Sidebar card · Release/Testing channel · Fallback notice

  • OS testing channel (Release / Testing)

    Organisations with access enabled can start new terminals with the Harness test version of the OS to check new functions early.

    Functions Release · Testing

  • Harness Preferences

    Personal settings relating to Harness: Workflow guide, onboarding tour, industry terms and BeeLens.

    Functions Workflow guide · Onboarding tour › Show again · Show industry aliases · BeeLens settings · Onboarding tour (Replay) · BeeLens

  • Connection and start notices (Connection lost, licence blocked, danic off, classic installation)

    BeeKeep3r IDE detects typical faults in the Harness connection and offers a concrete way out for each, rather than silently starting without Harness.

    Functions Harness connection lost · Harness commands locked · Claude not connected · Harness is turned off in Claude Code here · Classic Danic OS install found · Runtime licence notice

  • Initialize with Harness (/beehive-start)

    For a project without a delivery record, offers to open Claude with '/beehive-start' so that Harness sets up the project (Greenfield or Brownfield) together with the user.

    Functions Dialog after workspace creation · Start from Project Insights · Auto-refresh

  • Connect a ticket system (/connect-ticket-system)

    When a project is added, offers to set up a ticket connector (Notion, Jira, Linear, GitHub, GitLab, Azure DevOps) via the Harness command.

    Functions Agent selection · One-time

  • Record probe, baseline and collect flow

    Project Insights checks which artefacts of the delivery record are present and starts the matching Harness collect commands in a Claude terminal with one click.

    Functions Delivery record artefacts · Collect buttons · Execution · Automatic refresh · Blocked notices

Project Insights

The delivery dashboard of a single project, drawn from its Harness delivery record.

  • Project Insights — access, navigation and sample mode

    Frame of the project dashboard with section navigation in five groups, project name, status footer and an honest sample mode for projects without a record or without a licence.

    Functions Navigation · Status footer · Worktree banner · Remote notice · Sample mode · n/a instead of 0

  • Cockpit (KPI overview)

    Start page of Project Insights with the project's delivery KPIs over a selectable period, model mix and delivery flow.

    Functions KPI tiles · Delivery KPIs · Model Mix · Delivery flow · First steps · Unreadable log lines

  • Board (ticket board)

    The project's ticket board in four columns with flow metrics and a detail view per card; without a connected ticket system it is derived from the delivery record.

    Functions Columns · WHAT TO PICK UP · Flow panel · Freshness banner · PROVENANCE

  • Features (feature table with CSV export)

    Sortable and filterable table of all the project's features with phase, gates, costs and times, plus drill-down per feature.

    Functions Columns · Summary · Drill-down · CSV export

  • Test Coverage

    Test coverage and test health of the project overall and per category, including the history of earlier runs.

    Functions Overall · BY CATEGORY · Hints · PREVIOUS RUNS · Drill-down

  • Code Quality

    Code health assessment by rubric with overall grade, dimensions, surfaces and the biggest problem areas; plus test quality (mutation) and AI slop.

    Functions Overall · Dimensions · Surfaces · Top offenders · Hints / Rubric

  • Test quality (mutation) with gate integrity

    Answers 'Would the tests find a bug in the changed code?' with the mutation score of the changed code and provides evidence of whether the gate really checked the code.

    Functions Score vs. threshold · Surviving mutants · Evidence (gate integrity) · Changed lines without a mutant · History and tools

  • AI slop (AI slop detection)

    Shows whether the code carries typical residue of AI generation (stubs, swallowed errors and the like), with score, categories and rules.

    Functions AI-slop score · By category · Rules · Top findings · Calibration

  • Review Effectiveness

    Measures how well the OS review gate works: reviewer consensus, dismissed findings, dissent rounds, panel coverage and the cost/benefit of the panel.

    Functions Figures · Dismissed rate over time · Findings by reviewer and model · Dispositions · Panel ROI · External reviewers · Work items

  • Compliance (scorecard + evidence status)

    Regulatory self-assessment per category with target value and history, supplemented by the evidence status of each statement (evidenced, open, not checked).

    Functions Category tiles · Drill-down · Evidence per framework · Open statements · Stale sources · Verdict word violations

  • Security Audit (scorecard + Human approval gate)

    Security self-assessment per category with open Critical/High findings, plus an evaluation of the HITL gate that stops irreversible A3 actions by the agents.

    Functions Scorecard · Human approval gate · Decision list · Who may approve · Warnings · Approval instructions

  • ISMS (read-only in the project)

    Shows the organisation's ISMS in the project as read-only; it is maintained under Insights › ISMS.

    Functions Read-only view · Deep link

  • Retrospective (Learnings)

    Collects the project's retrospectives and learnings notes per feature and opens them for reading.

    Functions DOCUMENTS · Document view

  • Evolution

    Chronicle of the project's self-improvement from the feature learnings, with a maturity badge.

    Functions Timeline · Maturity level

  • ROI Metrics + BeeKeep3r Runtime runs

    List of all delivery runs with AI costs compared with a classic team equivalent, savings, speed-up and duration; plus the workflow runs of the BeeKeep3r Runtime.

    Functions Aggregate · all runs · Run list · Run detail · BeeKeep3r Runtime runs

  • Stubs (Wiring Radar)

    Finds unwired places in the project: confirmed stub markers with a feature reference and heuristic findings that a human has to assess.

    Functions Confirmed stubs · Heuristic findings

  • Project Insights Settings (ticket system, hourly rates, notifications)

    Project-specific settings that write the same files as the OS cockpit: ticket system (BYOT), role hourly rates, notifications and ISMS status.

    Functions Ticket system (BYOT) · Role hourly rates · Notifications · ISMS status · Read layer

Insights

Fleet view across all open projects: which project needs attention, and where.

  • Insights menu in the sidebar

    Expandable navigation to all Insights pages with red counter badges for projects in a critical state.

    Functions Sub-items · Badges

  • KPIs (fleet overview)

    Answers 'Which project needs attention first, and in which category?' with category tiles and a heatmap of projects × categories.

    Functions Categories · Projects × categories · Collect · Project drawer

  • Insights › Test Coverage

    Which projects are losing test protection, and where is it falling? Dimension tiles above a heatmap of projects × dimensions.

    Functions Dimensions · Heatmap

  • Insights › Code Quality

    Which project is slipping below the quality bar, and in which dimension? Including mutation score, AI slop and gate integrity.

    Functions Dimensions · Gate integrity · Top offenders / Collector hints

  • Insights › Review

    Where do reviewers agree, how much do they dismiss, and does every high-risk change get its full panel?

    Functions Review figures · Projects × review figures

  • Insights › Compliance

    Which regulatory family is below target, and in which project? Including sorting by open evidence.

    Functions Families · Open evidence

  • Insights › Security Audit

    Where are findings open, and how severe are they? Plus the HITL gate metrics of all projects.

    Functions Dimensions · A3 blocked · Policy overrides

  • Insights › Memory

    Which project memory is decaying through orphaned notes or silence?

    Functions Dimensions · Switch Memory hub

  • Insights thresholds

    Adjust the thresholds (OK/WATCH/CRIT) of the fleet view; changes take effect without a new Collect.

    Functions Thresholds · Restore defaults · 15 threshold values

BeeLens

Shows live what agents are doing during Harness commands, what it costs, where the project stands and under which rules the work was done.

  • BeeLens Run (live run view)

    Shows the running Harness command for the focused terminal: step progress, active agents with model and tokens, and the token count.

    Functions Step progress · Agents · Tokens · Recent runs · Other agent programs · Models · Focus

  • Token receipt

    Receipt for a completed run: where the tokens went, by phase, agent and model, optionally with a euro estimate and CSV export.

    Functions Breakdown · Ø this command · Euro value · Step sequence · Export CSV

  • Handoff card

    At the end of a run, shows where it stopped and which command makes sense next, and inserts it into the terminal on request.

    Functions Summary · Next command · Can be switched off

  • Questions (questions inbox)

    Collects questions that a run puts to the user and allows answers, which are inserted into the terminal as an '/answer-question' line.

    Functions Inbox · Live question · Notification

  • BeeLens Project (project position)

    Shows where the focused project stands in its lifecycle, which features are in which phase and which next steps make sense.

    Functions PROJECT LIFECYCLE · FEATURE MAP · SENSIBLE NEXT STEPS · Health

  • BeeLens Governance

    Under which rules did the agents work? Autonomy ladder A0–A3, guardrails of the run and gate ledger of a feature with audit export.

    Functions Autonomy ladder · Guardrails · Gate ledger · Audit export

  • Lens pill in the footer, hints and notifications

    BeeLens accompanies runs even when the panel is closed: footer pill with phase and tokens, first-time hint, run-end toast, notification when a run is waiting, and 'Since you were away'.

    Functions Footer pill · First-time hint · Run-end toast · Waiting notification · Since you were away · Sidebar badge

  • Insights › Lens (all projects)

    Which run is waiting for me or is stuck, and where does it stand? Overview of the BeeLens runs of all projects with a daily timeline.

    Functions Tiles · Project heatmap · Today timeline · Questions per project

  • BeeLens settings

    Switches BeeLens on/off and controls how intrusive it is.

    Functions BeeLens · Open panel on every run · Tokens in the footer · Handoff card at run end · Notify when a run waits for you

  • BeeLens demo (without Harness)

    Shows free users all three BeeLens tabs with sample data and offers a trial licence.

    Functions Demo data · Trial CTA

ISMS (ISO/IEC 27001)

Maintenance of the organisation's information security management system in its own Git project, directly from BeeKeep3r IDE.

  • Connect the organisation ISMS

    A Git project is connected as the central storage location for the ISO/IEC 27001 documentation; all projects read it, and it is maintained under Insights.

    Functions Connect · Status card · Change… / Disconnect · Problem card

  • ISMS Dashboard

    Maturity of the ISMS at a glance: releases, drafts, open placeholders, upcoming reviews, SoA and risk status.

    Functions Tiles · Next reviews · Suggestion

  • ISMS Maintain (create, edit and release documents)

    Document list of the ISMS with status; create documents from templates, import existing documents, fill placeholders and release with release metadata. Every write action shows a diff beforehand and is audited.

    Functions Document groups · Templates · Diff before writing · Release · Conflict detection · Audit log

  • ISMS editor (Form / Rich / Raw) and organisation profile

    Documents can be edited as a form based on placeholders, in a rich Markdown editor or raw; the organisation profile pre-fills placeholders.

    Functions Form · Rich · Raw · Organisation profile

  • Statement of Applicability (SoA editor)

    Assessment of the 93 Annex A controls: applicable yes/no, justification, implementation status and evidence.

    Functions Columns · Filter · Suggestions · Progress

  • Risk register (risk editor)

    Risk register with assessment, treatment, controls, owners and status; validates every row.

    Functions Columns · Validation · Summary

  • ISMS wizards

    Guided step-by-step wizards for building the ISMS, with saving of intermediate progress and resuming.

    Functions Start ISMS · Statement of Applicability · Capture risk · Adopt policy · Resume/Discard

Dependency Watch

Check the third-party dependencies of all projects against OSV, CISA KEV and EPSS, prioritise them and document them.

  • Insights › Dependencies

    Vulnerable packages of all open projects at a glance with 'Fix first' prioritisation.

    Functions Tiles · Heatmap · Fix first · Ecosystems · Trend

  • Project Insights › Dependencies

    Vulnerable packages of a project with filters, a licence column and a detail view including dependency path and suggested fix.

    Functions Table · Detail · Fix via Harness · Project KPIs

  • Not affected (VEX)

    Documents, with a justification, that a vulnerability does not affect the project and writes a VEX statement to vex.cdx.json.

    Functions Justification · Preview · Revoke · Not affected (n)

  • Release gate, Shipped product and CRA notice

    Shows the result of the dependency gate, marks a project as a shipped product and then displays CRA reporting deadlines for actively exploited vulnerabilities.

    Functions Release gate (deps-gate.json, DEPG-001) · Shipped product · CRA notice

  • Licences (licence policy display)

    Shows licence findings for the packages according to the licence policy (display only).

    Functions Findings

  • New packages (reputation gate)

    Shows the decisions of the OS hook that checks every new package before installation (BLOCK, ASK, WARN, ALLOW).

    Functions KPI cards · Decision list

  • Central daily check (BeeKeep3r SaaS)

    The package inventory (public purls) is checked centrally against vulnerability data every day; results appear alongside the local ones.

    Functions Origin display

Atlas

Shows how the open projects are structured and how they relate to each other; with Harness, agents explain the architecture.

  • Atlas Map

    A map of all projects open in BeeKeep3r IDE from a local scan (0 tokens, read-only), including calls that lead from one repository into another.

    Functions Fleet map · Scope · Drill-down · Technical · Refresh map · Legend · Languages

  • Atlas Full (Describe, Flows, Recent work, Health, Ask)

    Agents name and explain every part of the map, play through flows step by step, show what the last run changed, run a quick check for risks and answer questions about the code.

    Functions Describe · Flows · Recent work · Health · Ask Atlas · Harness session below the map · Auto-refresh

  • Atlas Export (HTML) and Export to repository (ARCHITECTURE.md)

    Shares the map as a single standalone HTML file or writes an ARCHITECTURE.md into a project — only on click, with a preview, never committed.

    Functions HTML export · Preview/diff · Safety

  • Atlas MCP for coding agents

    A local, read-only MCP server through which coding agents query the architecture instead of searching the repository; plus a short architecture brief in the server instructions.

    Functions atlas_overview · atlas_find · atlas_show · atlas_path · atlas_impact · atlas_flows · Architecture brief · Staleness notice

BeeKeep3r Harness

153 features · 21 areas

To the Harness

Delivery & access (BeeKeep3r Runtime / BeeKeep3r MCP)

How the BeeKeep3r Harness reaches the customer: as a signed OS pack that the BeeKeep3r Runtime delivers via MCP to Claude Code, Codex, Cursor, OpenCode or Gemini CLI, without OS files being read locally.

  • BeeKeep3r MCP tools

    The OS is available exclusively through the danic tools of the MCP server: catalogue, commands, skills, agents, scripts, knowledge, workflows, gates and rules. This keeps the OS content protected and identical everywhere.

    Functions danic_catalog · danic_command · danic_skill · danic_agent · danic_script · danic_read · danic_workflow_start / next / submit / status · danic_workflow_rewind · danic_gate · danic_status · danic_rules

  • Workflow engine with gates

    Eleven core commands run on the Runtime as a state machine: steps are executed in sequence, gates block deterministically, and path outputs are checked against the repo. This prevents phases from being skipped.

    Functions Workflow commands · Step types · Path contract · Context for reviewers · Sealed scoring

  • OS versions, OS pack & channels

    The OS is delivered as a versioned, signed and encrypted OS pack (currently 8.0.0, channel stable). DANIC can set a minimum OS version with a deadline per channel; internal content is excluded from the customer pack by a deny list.

    Functions Version manifest · Channels · Minimum OS per channel · Pack signature & payload key · Customer pack filter

  • Content protection for OS prompts

    Agent definitions are delivered protected by default, scoring models are sealed, and the Prompt Confidentiality rule prohibits verbatim reproduction or export of OS texts.

    Functions Classification · Sealed knowledge files · Prompt Confidentiality

BeeFlows: Start Greenfield: specification & product setup

From the initial idea to a validated specification and the technical foundation of a new product.

  • /start — Start

    Entry point for every project: checks the prerequisites (Node, git, OS package), detects greenfield or brownfield and idempotently creates the delivery structure delivery/.

    Functions Preflight · Greenfield/brownfield detection · Delivery scaffold · ID scheme explanation

  • /spec-interview — Domain interview

    Guided, example-based interview with a domain expert that fills the glossary, workflow map, rule catalogue, decision log and domain debt log with source references. Does not invent facts; unknowns are recorded as domain debt.

    Functions Onboarding · Elicitation funnel · Incremental domain artefacts · Domain Debt · Stop criteria

  • /validate-spec — Validate spec

    Spec Quality Gate: assesses the specification (InputSpec.md or spec bundle) for completeness, unambiguity and compliance coverage and writes a spec quality report with a score and gate decision. Mandatory before /init-product-system.

    Functions Resolve target & drift check · Structure score · Quality scoring · Gate decision

  • /ingest-spec — Documents to specs

    Converts existing customer specifications into draft spec bundles that /validate-spec can assess. Detects OpenSpec change sets (deterministic, repeatable import, OpenSpec remains the source of truth) or arbitrary documents such as PDF, DOCX and Markdown.

    Functions Format detection · OpenSpec import · Document clustering

  • /refine-spec — Refine spec

    Iterative spec quality coach: reads the latest quality report, proposes a concrete correction for each gap and applies it after confirmation, until the target score is reached or an override is documented.

    Functions Refinement briefing · Set target score · Gap loop · Re-validation

  • /answer-question — Answer open questions

    Works the Product Owner's answers to open questions (OPEN rows in open-questions.md or HALTED-ASKED rows in deviations.md) into the delivery record; every change is shown as a diff and applied only after confirmation.

    Functions Resolve references · Incorporate answers · Update status

  • /design-system — Design system

    Brand pack onboarding in two modes: ingest takes over existing design sources (Figma, website URL, assets, CI PDF, code), interview asks for preferences and synthesises design tokens with a live preview. The result is a canonical brand pack whose compliance is enforced by a design gate.

    Functions ingest mode · interview mode · Write brand pack · Design gate

  • /init-product-system — Set up product

    Builds the product foundation from a validated specification: PRD, architecture baseline, ADRs and further baselines. Runs once per product.

    Functions Check spec prerequisite · Input qualification · Product framing · Architecture & ADRs

  • /init-feature-foundation — Technical foundation

    Sets up the technical foundation: repo and project structure, tech stack decision, CI/CD and test setup. The foundation readiness check then releases feature delivery.

    Functions Context assessment · Repo and project structure · CI/CD and test baseline

  • /estimate-effort — Estimate effort

    Threefold effort and cost estimate for a scope: classic team, AI-assisted team and danic agentic setup. Writes a machine-readable estimate as the single source and renders an HTML analysis and a draft quote from it.

    Functions Break down scope · Calculate three modes · Write estimate · Report & quote

  • /ia-spec — Information architecture

    Structures the information architecture of a feature before the UX specification: screen and route inventory, OOUX object model, sitemap, navigation model and labelling. Documents gate IAG-001.

    Functions Read design profile · Apply IA method · Write ia-spec.md · Gate IAG-001

BeeFlows: Start Brownfield: taking over existing systems

Onboard existing codebases for agentic delivery without putting the existing system at risk.

  • /init-legacy-system — Onboard existing system

    Reverse-engineers an existing codebase into a deliverable foundation and sets the evolution mode: modernize, preserve or hybrid. Works read-only on the existing code.

    Functions Intake & mode selection · Complexity class · Codebase ingestion · Reverse ADRs · Mode-dependent analysis & handoff

  • /extract-design-system — Extract design

    Reverse-engineers a design system from a live URL, screenshots or frontend code into verifiable tokens (tokens.json, tokens.css, DESIGN-SYSTEM.md, token-sheet.html), optionally with adoption into the brand pack.

    Functions Clarify source · Extraction · Write artefacts · Confidence & gaps

  • /wrap-system-cli — CLI harness

    Makes an existing system operable for agents by generating a lean command-line interface to a real backend (no reimplementation). Currently local subprocess backends only.

    Functions Analysis & design · Implementation & tests · Documentation & publication

BeeFlows: Feature delivery

Deliver features end to end through phases and gates, at varying levels of rigour.

  • /ship-feature-auto — Ship feature

    The standard path for a feature: delivers it end to end through all phases (run forecast, intake, framing, architecture, implementation, review with dissent round, tests, release) with documented gates and an up-to-date delivery record.

    Functions Run forecast · Intake & framing · Implementation · Review panel & dissent · Partial runs · Brownfield profile

  • /ship-loop — Ship feature strictly

    Delivers a feature with hard anti-shortcut gates: it only counts as delivered once the logic is reachable in the running app, all guards are green and an E2E test passes in a real environment.

    Functions Read spec & forecast · Logic & wiring · Guards green · Real-environment E2E

  • /ship-feature-light — Ship feature light

    Light delivery mode for small, low-risk features (UI, meta or docs only) with reduced gate effort.

    Functions Suitability check · Intake & framing · Reduced gates

  • /ship-feature-tests — Ship feature with tests

    Implements automated tests for a feature based on the test strategy and test cases and integrates them into the CI pipeline.

    Functions E2E implementation · CI integration · Completion

  • /plan — Plan implementation

    Restates the requirements, assesses risks and creates a step-by-step implementation plan; does not touch any code before the user confirms.

    Functions Restate requirements · Phases & dependencies · Risks & complexity · Wait for confirmation

  • /checkpoint — Checkpoint

    Saves an interim delivery state, e.g. before risky steps or at the end of a block of work.

    Functions Write checkpoint

  • /spec-retrospective — Spec retrospective

    After delivery, checks how closely what was delivered matches the specification and produces a spec adherence report with a score and drift analysis.

    Functions Adherence check · Adherence report

BeeFlows: Tests & Quality Gates

Test strategy, test cases, test execution and the review gates of a delivery.

  • /gate-review — Technical review

    Technical Review Gate: an independent review agent (Maker≠Checker) reviews the implementation and writes findings JSON as well as the REV evidence required by the release gate check.

    Functions Technical review · Review report & gate decision · Blind dissent round

  • /gate-qa — QA review

    QA Gate: the QA manual and QA automation agents test the feature; the Gate 5 decision is then made and status.md is updated.

    Functions Manual QA · Automated QA · Gate 5

  • /quality-gate — Quality check

    Generic quality check outside the standard phase chain: detects language and tooling, runs formatter, lint and type checks and returns a list of remediation items.

    Functions Detect tooling · Format, lint and type checks · Remediation list

  • /tdd — Test-driven development

    Enforces test-driven development: define interfaces, write failing tests first, implement minimally, refactor and demonstrate at least 80 % coverage.

    Functions Interfaces · RED · GREEN · Refactor & coverage

  • /e2e — End-to-end tests

    Generates and runs end-to-end tests with Playwright: user journeys, cross-browser runs, screenshots, videos and traces as artefacts.

    Functions Analyse user flow · Generate Playwright tests · Run & capture · Report & flaky detection

  • /test-feature-spec — Test cases from spec

    Creates the functional test cases and the technical test architecture of a feature; two agents work in parallel.

    Functions Test design · Test architecture

  • /test-feature-strategy — Test strategy

    Creates a test strategy for a feature: analyse risks, recommend test levels, define test scope.

    Functions Identify risks · Test levels · Strategy document

  • /test-suite-review — Review test suites

    Reviews the test suite for quality, flakiness, maintainability and coverage gaps; the test review and coverage traceability agents work in parallel.

    Functions Test review · Coverage & traceability

  • /perf — Performance

    Performance as measured evidence: budgets per operation (p50/p95/p99, throughput, error rate), k6 load tests with thresholds and profiling of slow operations. Never against production.

    Functions budget mode · load mode · profile mode

  • /go-test — Test Go

    TDD workflow for Go: table-driven tests first, then implementation, coverage ≥ 80 % with go test -cover.

    Functions Types & interfaces · Table-driven tests · Implementation & refactor · Coverage

BeeFlows: Release & operations

Release approval, migrations, deployment and observability as evidenced, human-confirmed steps.

  • /gate-release — Release review

    Release Gate: resolves the feature folder, runs the deterministic release gate check and assesses release readiness through to go/no-go.

    Functions Deterministic release gate check · Release readiness · Go/No-Go

  • /db-migrate — Migrate database

    Database migrations with proof: forward and rollback pair from a schema change, independent review, dry run with timing and locks, and a rollback rehearsal. Never against production.

    Functions Classify · author · review · dry-run

  • /deploy — Deploy

    Deployment and rollback as a traceable, human-confirmed action: plan with runbook, execution only after confirmation per environment (production only with a green release gate), data-safe rollback.

    Functions plan · execute · rollback

  • /observability — Observability

    Observability as a deliverable: stack selection and instrumentation, SLOs with error budgets and alert rules, or a review of existing logs, traces and metrics.

    Functions setup · slo · review

BeeFlows: Bugs, Hotfixes & Incidents

Fix defects in a planned way or in an emergency, repair builds and work through incidents up to the postmortem.

  • /fix-bug — Fix bug

    Planned bug fixing via a reduced delivery flow: find the cause, create a bug record, start with a failing test, fix, verify and close.

    Functions Run forecast · Intake & root cause · Fix · Review & completion

  • /hotfix — Hotfix

    Emergency fix when production is broken for users: triage, fix, regression test, test gate, review, deployment and a subsequent audit.

    Functions Triage · Fix & regression test · Review & deployment · Post-hoc audit

  • /build-fix — Repair build

    Fixes build and type errors: detects the build system, groups errors and fixes them one at a time.

    Functions Detect build system · Group errors · Fix loop · Completion

  • /go-build — Build Go

    Fixes Go build errors, go vet warnings and linter messages incrementally and minimally via the go-build-resolver.

    Functions Diagnosis · Incremental fixes

  • /investigate-incident — Investigate incident

    Structured incident investigation through to root cause analysis (RCA). Excluded for clinical chains.

    Functions Evaluate symptoms · Investigation · RCA

  • /generate-postmortem — Draft postmortem

    Generates a DRAFT postmortem from the most recent RCA (timeline to prevention).

    Functions Load RCA · Postmortem draft

BeeFlows: Review & Pull Requests

Independent code reviews, refactoring, verification and pull request workflow.

  • /code-review — Code review

    Independent review of the pending diff by a reviewer agent, stack-aware and with machine-readable findings; blocks on CRITICAL/HIGH.

    Functions Stack detection · Checks · Findings report

  • /refactor-clean — Remove dead code

    Finds and removes dead code and consolidates duplicates in a safe deletion loop.

    Functions Detect & categorise · Safe deletion loop · Consolidate duplicates

  • /refactor-7track — Refactoring in seven passes

    Orchestrated clean-up pass as a quality gate with seven specialised sub-agents.

    Functions Track 1–2 · Track 3–4 · Track 5–6 · Track 7

  • /verify — Verify

    Verification check after a change: build, types, lint, test suite, console.log audit and git status.

    Functions Build & types · Lint & tests · Audit

  • /pr — Pull Request

    Pull request workflow based on the delivery record: create a PR, review other people's PRs or work through review comments; every write to the remote happens only after confirmation.

    Functions create · review · Address comments

BeeFlows: Security

Security review per feature, per diff, for dependencies and as a coordinated sweep before release.

  • /security-check-feature — Check security

    Security check for a feature: risk classification, threat model by the security architect and gate decision.

    Functions Risk classification · Threat model · Gate decision

  • /security-review-diff — Review diff for security

    Semantic, high-confidence security review of the branch changes following the methodology of Anthropic's claude-code-security-review, extended with multi-tenancy, GDPR and AI categories.

    Functions Find vulnerabilities · Filter false positives · Report

  • /scan-dependencies — Scan dependencies

    Scans the dependencies for known vulnerabilities, with fix suggestions.

    Functions Detect project type · Run scanners · Classify & fixes

  • /deps-upgrade — Upgrade dependencies

    Dependency upgrades with proof: plan from the scanner, batch-by-batch application on a chore branch, /verify after each batch, automatic revert of red batches and a re-scan as evidence that the CVEs are closed.

    Functions Scan & plan · Batches · Revert or keep · Re-scan

  • /security-pre-release-sweep — Security before release

    Coordinated security sweep before release: SAST, dependencies, DAST, RLS audit, response leaks, configuration/headers, AI red team and aggregation.

    Functions SAST & dependencies · DAST · RLS, leaks, headers · AI red team · Attack simulation · Aggregation

  • /security-report — Security report

    Security readiness report: collects security artefacts, aggregates findings, calculates a risk score and maps findings to compliance frameworks.

    Functions Collect & aggregate artefacts · Risk score · Executive Summary · Compliance mapping

BeeFlows: Compliance (AIUC-1, CRA, ISO 27001)

Regulatory audits and ISMS documentation for regulated products (healthcare, EU).

  • /audit-aiuc — Compliance audit (AIUC-1)

    AIUC-1 compliance audit against the state of the repo: coverage matrix across all six control domains A–F, gap list and prioritised recommendations; clinical agents automatically receive a CRITICAL flag.

    Functions Discovery & clinical flag · Coverage matrix · Clinical section · Gap list & crosswalks

  • /iso-doc — ISO documentation

    Guided self-service documentation of an ISMS in line with ISO/IEC 27001:2022: scope, Statement of Applicability (93 Annex A controls), risk methodology and mandatory clause evidence, by interview or as a draft from repo signals.

    Functions interview mode · draft mode · status mode

BeeFlows: Design, SEO, documentation & Atlas

Discoverability (SEO/GEO), system documentation and the agentic architecture map Atlas Full.

  • /seo — SEO & GEO

    SEO and Generative Engine Optimization in one command: audit, design-time advice and safeguarded fixes; never invents evidence.

    Functions audit · consult · fix

  • /generate-docs — Generate system diagrams

    Generates or extends the four canonical Mermaid system diagrams (C4 context, ER, user journeys, infrastructure topology) exclusively from repo sources; if a source is missing, the diagram is skipped with a stated reason.

    Functions Status · Collect sources · Generate diagrams

  • /docs — Maintain documentation

    Documentation maintenance in one command: synchronise generated docs from source-of-truth files or generate token-lean codemaps.

    Functions sync · codemaps · Diff & staleness

  • /graph-os — Knowledge graph

    Builds or updates a knowledge graph with graphify (interactive visualisation plus GraphRAG JSON).

    Functions Generate graph

  • /atlas-describe — Atlas: Describe

    Atlas Full: an agent names and explains every part of the cross-project architecture map and writes the structure, descriptions, key flows and an introductory tour to the Atlas folder of the BeeKeep3r IDE; read-only on the projects.

    Functions Load Atlas input · Worksheets · Structure & validation · Descriptions, flows, tour

  • /atlas-flow — Atlas: Flow

    Atlas Full: traces a requested flow (e.g. “How does a refund work?”) across the map and writes it as steps that the BeeKeep3r IDE plays back on the map; read-only on the projects.

    Functions Trace flow · Write flow

  • /atlas-ask — Atlas: Ask

    Atlas Full: answers an architecture question about the project fleet read-only using find, show and path across the map; the answer board shows the boxes involved.

    Functions Investigate · Write answer

BeeFlows: Status, metrics & Insights data collection

Project and feature status, leadership metrics and the collect commands that populate the Insights dashboards.

  • /feature-status — Feature status

    Shows the current status of a feature from the delivery record.

    Functions Show status

  • /team-status — Team status

    Shows all teams, active features and cross-team dependencies.

    Functions Team registry · Assign features · Dependencies & release coordination

  • /metrics — Metrics

    Leadership report with lead time, costs and token efficiency; updates the metrics dashboard and exports the data for the Insights views.

    Functions Calculate metrics · Update dashboard · Export · Feature view

  • /visualize — Visualise metrics

    Generates KPI charts from the metrics dashboards, directly in the chat or as dashboard files.

    Functions Read & check data · Generate charts · Write dashboards

  • /collect-all — Collect everything

    Updates all Insights artefacts in one idempotent, fault-tolerant run and reports written, unchanged, skipped or failed for each step.

    Functions Scaffold · Coverage & code quality · Compliance, security, dependencies · Review metrics & overview · Ticket board

  • /collect-test-coverage — Collect test coverage

    Collects unit, frontend and E2E suites into quality/test-coverage.json for the test coverage dashboard; E2E is measured as pass rate and flow coverage.

    Functions Discovery & run · Normalise · --gaps

  • /collect-code-quality — Collect code quality

    Collects ESLint, tsc, ruff and jscpd into quality/code-quality.json for the code quality dashboard; deterministic score across seven dimensions including ai_slop. Measures, but never fixes.

    Functions Discovery & run · Score · Mutation & slop

  • /collect-compliance — Collect compliance

    Updates the project's compliance scorecard and writes a history point stamped with the Git SHA.

    Functions Aggregate scorecard · Report scores

  • /collect-security — Collect security

    Updates the project's security scorecard and writes a history point; final step of /security-pre-release-sweep.

    Functions Aggregate scorecard · Report scores

BeeFlows: Tickets & Board (BOYT)

Connect your own ticket system (Bring Your Own Tickets) and transfer tickets into delivery.

  • /connect-ticket-system — Connect ticket system

    Guided setup of the BOYT connector: provider, board, actor, status mapping and token. Writes .beehive-ticketing.yaml to the repo root and stores the token in a gitignored sidecar file, never in the repo.

    Functions Choose provider · Board, actor, status · Store token securely · Check wiring

  • /ingest-tickets — Ingest tickets

    One inbound poll cycle: watches the configured boards, picks up ready tickets, generates a spec bundle from them and writes the status back. A no-op without configuration.

    Functions Check configuration · Load adapter · Cycle

  • /collect-ticket-board — Collect ticket board

    Reads the connected ticket system read-only via its MCP server and normalises it to tickets/ticket-board.json with the four columns backlog, doing, review and released.

    Functions Writeback drain · Read & normalise · Output & check

BeeFlows: Learn & Improve

Capture and apply learnings, run retrospectives and improve agents/skills in a targeted way.

  • /learn — Learnings

    Extracts reusable patterns from the session, scores them against a rubric and, after confirmation, stores them globally or per project.

    Functions Review session · Determine storage location · Self-assessment · Confirmation

  • /apply-learnings — Apply learnings

    Reads the learnings collected from earlier features and feeds them into the current delivery context.

    Functions Collect & filter · Ranking · Brief & injection

  • /retro — Retrospective

    Retrospective with the user after a feature delivery: aggregates agent signals, asks a fixed sequence of questions, synthesises retro.md and stores confirmed learnings as proposals (propose-only).

    Functions Aggregate · Interview · Synthesis

  • /retro-approve — Approve proposal

    The only way to activate an instance overlay proposal: four-eyes principle with initials and timestamp, diff view and approval marker; no auto-commit.

    Functions Show diff · Activate · Approval marker

  • /improve-skills — Improve skills

    Skill Evolution Engine: checks the health, freshness and usage of the skills and recommends FIX, DERIVE or CAPTURE.

    Functions Freshness & usage · Decision model · Queue & Ranking

  • /run-experiment — Run experiment

    Runs an A/B experiment, e.g. two prompt or agent variants, and evaluates the result.

    Functions Design · Execution · Evaluation

BeeFlows: Skills & OS Quality

Create, validate and diagnose skills, check rule adherence with trap tests, update the OS.

  • /audit-skills — Audit skills

    Static conformance check of the YAML frontmatter of all skills and agents against the agentskills.io standard; purely local, deterministic, without an LLM.

    Functions Discovery & Parse · Apply rules · Report

  • /skill-create — Skill from history

    Analyses the local Git history, detects recurring coding patterns and generates SKILL.md files from them.

    Functions Discovery check · Detect patterns · Generate SKILL.md

  • /skill-validate — Validate skill

    Produces a held-out PASS report for a skill, bound to the SHA: disjoint test cases run with a fresh subagent; the gate reads only the artefact.

    Functions Check disjointness · Run cases · Hash & Report

  • /skill-doctor — Skill diagnosis

    Diagnoses unused and orphaned skills based on the skill invocation log (warnings only).

    Functions Unused/orphan report

  • /harness-audit — Audit Harness

    Audits the configuration of the agent Harness and delivers a scorecard with actions.

    Functions Inspect Harness · Score categories · Actions

BeeFlows: Session & Navigation (incl. NL Layer)

Daily working rhythm, command navigation and free-text control of the OS.

  • /nl — Free-text request

    Translates natural language into delivery actions: detects the intent, reads the project state and maps it to an existing command, agent or workflow.

    Functions Detect intent · Project state · Mapping · Clarifying question

  • /help — Find commands

    Command navigator: shows matching commands for a role, a delivery phase or a single command, read from catalog.json and never invented.

    Functions Quick Access · Role and phase view · Single command

  • /daily — Daily status

    Daily standup: loads the context and shows priorities; optionally with a daily note.

    Functions Load context · Status overview · Daily note

  • /tldr — Close session

    Session summary: records decisions, updates memory.md, extracts learnings, and maintains the operator context and the feature status.

    Functions Session analysis · memory.md · Learnings & Brain · Feature status

BeeFlows: Autonomy & Scaling (incl. Headless)

Autonomous multi-feature runs, headless delivery, cost modes and multi-model work for advanced users.

  • /ship-product-auto — Ship product

    Autonomous end-to-end loop from the InputSpec to finished, hardened software: validate and sharpen the spec, build the foundation, deliver all features via /ship-feature-auto, apply learnings and run several polish-and-harden cycles until the stop conditions are reached.

    Functions Spec & Foundation · All features · Polish & Harden · Status

  • /advisor — Architecture advice

    In planner/executor mode, escalates specifically to the planner (Opus tier, highest thinking level) and delivers an addendum to the PlanArtifact.

    Functions Escalation · PlanArtifact addendum

  • /multi — Multi-model

    Multi-model collaboration with Codex (backend authority) and Gemini (frontend authority) via a wrapper; Claude retains sovereignty over the code.

    Functions plan · execute · backend / frontend · workflow

  • /model-route — Choose model

    Model routing by task class and data class (guardrail G-24): data of class Restricted is never given a public cloud model.

    Functions Show/set routing · Data class protection

  • /headless-run — Start run

    Starts an autonomous, spec-driven delivery via the Headless Conductor: preflight, execution graph (DAG), execution via adapters and a concluding final-report. Never asks questions during the run.

    Functions Preflight · Build DAG · Adapter execution · final-report

  • /headless-resume — Resume run

    Resumes an aborted or resumable headless session from the last successful checkpoint; checks branch, state and spec hash, and work that has already been paid for is not repeated.

    Functions Verification · Resume

  • /headless-status — Run status

    Shows the live state of a headless session in compact form; read-only.

    Functions Read status

  • /headless-abort — Stop run

    Ends a running headless session in an orderly way: an abort signal is checked between nodes, never in the middle of a node.

    Functions Abort signal

  • /headless-retro — Run review

    After a headless session, creates the cross-feature retro summary from the retro index collected during the run; pure aggregation, proposals only.

    Functions Aggregation

Bees (Agent Catalogue, 128 Agents)

The specialised AI agents of the OS, called “Bees” in customer language, grouped by area of responsibility.

  • Bees: Orchestrators (7)

    Coordinate all work; no agent runs without orchestrator routing. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions ops-orchestrator-agent · foundation-orchestrator-agent · foundation-legacy-orchestrator-agent · delivery-orchestrator-agent · nl-orchestrator-agent · planner-executor-orchestrator-agent · cli-harness-orchestrator-agent

  • Bees: Product & Planning (10)

    Problem space, specification, planning and estimation. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions product-manager-agent · product-owner-agent · spec-steward-agent · spec-refinement-agent · delivery-planner-agent · implementation-planner-agent · effort-estimator-agent · planner · planner-advisor-agent · chief-of-staff

  • Bees: Architecture (3)

    Solution and software architecture. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions solution-architect-agent · software-architect-agent · architect

  • Bees: Stack Developers Frontend (7)

    Frontend specialists that are activated automatically per tech stack. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions frontend-developer-agent · frontend-design-agent · react-typescript-developer-agent · nextjs-developer-agent · vue-developer-agent · angular-developer-agent · blazor-developer-agent

  • Bees: Stack Developers Backend (6)

    Backend specialists per tech stack. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions backend-developer-agent · spring-boot-developer-agent · dotnet-developer-agent · nestjs-developer-agent · fastapi-developer-agent · go-developer-agent

  • Bees: Stack Developers Mobile (4)

    Mobile and cross-platform specialists. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions flutter-developer-agent · kotlin-android-developer-agent · swiftui-developer-agent · maui-developer-agent

  • Bees: Infrastructure & Operations (7)

    CI/CD, platform, observability and performance. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions devops-agent · kubernetes-agent · terraform-agent · observability-agent · performance-engineer-agent · messaging-agent · networking-agent

  • Bees: Stack Reviewers (7)

    Stack-specific code reviewers for /code-review --stack auto. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions react-typescript-reviewer · spring-boot-reviewer · dotnet-reviewer · blazor-reviewer · maui-reviewer · go-reviewer · python-reviewer

  • Bees: Quality & Review (16)

    Independent review, QA and refactoring. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions review-agent · dissent-critic-agent · review-consolidator-agent · code-reviewer · qa-automation-agent · qa-manual-agent · e2e-runner · tdd-guide · refactor-7track-orchestrator · refactor-7track-deduplication · refactor-7track-type-consolidation · refactor-7track-dead-code · refactor-7track-circular-deps · refactor-7track-type-strength · refactor-7track-error-handling · refactor-7track-legacy-slop

  • Bees: Test Specialists & Collectors (13)

    Test architecture, test authoring and the collector agents for the Insights data. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions test-architect-agent · test-design-agent · test-review-agent · test-strategy-agent · playwright-architect-agent · coverage-traceability-agent · ci-test-agent · frontend-test-author-agent · backend-test-author-agent · test-coverage-collector-agent · board-dispatch-agent · code-quality-collector-agent · ticket-board-collector-agent

  • Bees: Security (10)

    Security design, scans and attack testing. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions security-architect-agent · security-reviewer · security-testing-agent · security-report-agent · sast-scanner-agent · dependency-scanner-agent · ai-red-team-agent · rls-auditor-agent · response-leak-detector-agent · config-headers-auditor-agent

  • Bees: Compliance, Accessibility & Discoverability (5)

    Regulation, WCAG and SEO/GEO. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions accessibility-agent · seo-agent · geo-agent · german-market-compliance-agent · aiuc-monitor

  • Bees: Database & Data (2)

    Database review and visualisation. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions database-reviewer · data-visualization-agent

  • Bees: Documentation, UX & Release (6)

    Documentation, information architecture, UX and release approval. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions documentation-agent · doc-updater · documentation-linking-agent · release-manager-agent · information-architecture-agent · ui-ux-agent

  • Bees: Bugfix, Build & Legacy (8)

    Bug fixing, build repair and legacy code analysis. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions bug-fix-agent · build-error-resolver · go-build-resolver · dotnet-build-resolver · legacy-code-analyzer-agent · legacy-complexity-classifier-agent · refactor-cleaner · incident-investigator

  • Bees: Meta & System Improvement (15)

    Learning, metrics, optimisation and experiments. The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions observer-learning-agent · system-improvement-agent · evolution-engine-agent · skill-librarian-agent · metrics-agent · loop-operator · harness-optimizer · workflow-optimizer-agent · routing-optimizer-agent · prompt-optimizer-agent · experiment-agent · experiment-evaluator-agent · human-benchmark-agent · spec-quality-agent · retrospective-facilitator-agent

  • Bees: SAP Track (2)

    First SAP building blocks (stubs without an SAP connection). The agents are deployed automatically by the orchestrators and commands; the OS activates stack specialists based on delivery/management/tech-stack.md.

    Functions sap-cap-engineer · sap-adt-bridge

Skills (92 Skills)

Method building blocks that agents and commands load on demand.

  • Skills: Development Workflow (10)

    Reusable method knowledge from the group “Development Workflow”, which agents and commands load on demand.

    Functions brainstorming · writing-plans · effort-estimation-skill · executing-plans · subagent-driven-development · dispatching-parallel-agents · using-git-worktrees · finishing-a-development-branch · cli-harness-generation-skill · think-in-code-skill

  • Skills: Quality & Review (13)

    Reusable method knowledge from the group “Quality & Review”, which agents and commands load on demand.

    Functions verification-before-completion · adversarial-verification-skill · spec-completeness-checker · spec-state-contract · scope-guard · ac-generator · guided-elicitation · skill-conformity-check · requesting-code-review · receiving-code-review · systematic-debugging · code-quality-audit-skill · refactor-7track

  • Skills: Testing (5)

    Reusable method knowledge from the group “Testing”, which agents and commands load on demand.

    Functions test-driven-development · test-strategy-skill · test-data-strategy-skill · coverage-enforcement-skill · trap-testing

  • Skills: Architecture & Design (23)

    Reusable method knowledge from the group “Architecture & Design”, which agents and commands load on demand.

    Functions adr-protocol-skill · adr-validation-skill · api-contract-skill · architecture-patterns-skill · nfr-framework-skill · nfr-guard · arch-guard · adr-conformance-check · frontend-design-skill · information-architecture-skill · ui-ux-pro-max-skill · design-ingest-figma-skill · design-ingest-vision-skill · ci-pdf-extractor-skill · extract-design-system-skill · design-token-synthesis-skill · interaction-motion-craft-skill · observability-stack-skill · database-migrations-skill · legacy-mode-profile-skill · performance-testing-skill · event-driven-implementation-skill · edge-patterns-skill

  • Skills: Security (6)

    Reusable method knowledge from the group “Security”, which agents and commands load on demand.

    Functions security-review-skill · threat-modeling-skill · rls-audit-skill · response-leak-detection-skill · security-headers-audit-skill · security-design-classification-skill

  • Skills: Compliance, Accessibility & SEO (11)

    Reusable method knowledge from the group “Compliance, Accessibility & SEO”, which agents and commands load on demand.

    Functions gdpr-compliance-skill · gdpr-scan · accessibility-audit-skill · seo-audit-skill · geo-audit-skill · geo-no-fabrication-skill · aiuc-compliance · eu-ai-act-classification-skill · iso-27001-doc-skill · cra-compliance-skill · mermaid-system-docs-skill

  • Skills: Delivery & Gates (9)

    Reusable method knowledge from the group “Delivery & Gates”, which agents and commands load on demand.

    Functions delivery-orchestration-skill · quality-gate-skill · gate-handoff-protocol-skill · release-readiness-skill · risk-assessment-skill · specialist-trigger-decision-skill · rca-formatter · postmortem-generator · ticket-integration-skill

  • Skills: System Improvement & Meta (9)

    Reusable method knowledge from the group “System Improvement & Meta”, which agents and commands load on demand.

    Functions skill-evolution-advisor-skill · evolution-decision-skill · writing-skills · using-superpowers · spec-quality-validation · context-budget-controller-skill · caveman-dev-skill · graphify · conditional-activation-patterns

  • Skills: Language Patterns & Testing (4)

    Reusable method knowledge from the group “Language Patterns & Testing”, which agents and commands load on demand.

    Functions golang-patterns · golang-testing · python-patterns · python-testing

  • Skills: SAP Track (2)

    Reusable method knowledge from the group “SAP Track”, which agents and commands load on demand.

    Functions sap-readiness-assessment · abap-quality-gate

Rules, Gates, Guardrails & Hooks

The guardrails that make agentic delivery reliable and auditable.

  • Always-on Rules (26)

    26 binding rules that apply in every Harness session and define the behaviour of the agents: delivery process, anti-shortcut, security, testing, NL routing and confidentiality.

    Functions agentic-delivery · anti-shortcut-delivery · agents · behavioral-constraints · clean-code · coding-style · context-management · customer-artifact-language · development-workflow · frontend-design · git-workflow · hooks · nfr-awareness · nl-orchestration · nl-state-awareness · patterns · performance · planner-executor-mode · prompt-confidentiality · secure-coding · security · task-modes · testing · ui-ux-pro-max · usage-audit-exclusions · visualization-standards

  • Quality Gates (9 Phase Gates)

    Nine gates at the phase transitions ensure that no feature moves on without a documented approval; hooks enforce the order.

    Functions Gate 1: Discovery Complete · Gate 2: Architecture Approved · Gate 3: Definition of Ready · Gate 4: Development Complete · Gate 5: QA Passed · Gate 6: UX Accepted · Gate 7: Product Accepted · Gate 8: Release Ready · Gate 9: Done · Foundation Readiness · Reduced gate set for bug & hotfix

  • Special Gates (IDs)

    Additional, deterministically checked gates for spec quality, design, accessibility, dependencies, test quality, AI regulation and human approvals.

    Functions SQVS-001 Spec Quality · IAG-001 Information architecture · TQG-001 Ticket quality · Discoverability Gate · Design Quality Gate · G10 Accessibility Gate · ADVG-001 Dissent · DEPG-001 Dependency gate · MUTG-001 Mutation gate · SLOPG-001 AI slop gate · AIAC-001 EU AI Act · HITLG-001 Human Approval · IDDG-001 ID duplicates · EST-001 Estimate consistency · ICVG-001 Anti-stub · SKVG-001 Skill validation · TRPG-001 Trap tests

  • Guardrails G-01 to G-26

    Architecture and security guardrails that agents follow when making decisions; several are technically enforced by hook.

    Functions G-01 Monolith first · G-02 Bounded context before a new service · G-03 No async without business need · G-04 Observability is not optional · G-05 No premature abstraction · G-06 Database first, cache later · G-07 Complexity Budget · G-08 Security and compliance are not optional · G-09 No gold plating · G-10 External before internal · G-11 Design Consistency · G-15 Risk Gate Enforcement · G-16 Legacy Safety Net · G-17 Tool Concurrency Safety · G-18 Provenance · G-19 Credential isolation · G-20 Lethal Trifecta / Default-Deny-Egress · G-21 HITL before destructive/irreversible · G-22 Tool surface budget · G-23 Behavioral Constraints · G-24 Data class cannot be overridden in the router · G-25 Per-loop budget · G-26 No skill merge without held-out

  • Permission Model A0–A3 & HITL Approval

    Every action falls into one of four classes: A0 read (automatic), A1 locally reversible, A2 substantive (declaration of intent) and A3 irreversible (only with human approval). Deployments, IaC, migrations and MCP calls are governed separately.

    Functions A0 Observe · A1 Local reversible · A2 Substantive · A3 Irreversible · Deploy/migration approval

  • Review-Gated Mode (Push Gate)

    Optional working mode (Track 2) in which git push is blocked until a human has approved the task manifest; the manifest and the approval remain in the repo as an audit trail.

    Functions Task manifest · Approval marker · Pre-push gate

  • Trap Tests (35 Regression Traps)

    Known-answer regression tests check whether agents really comply with the rules (e.g. percentage and margin traps, ticket injection, MCP attacks, memory poisoning, Atlas-First). Any real failure can become a permanent trap.

    Functions Calculation and assumption traps · Compliance and process traps · Security traps · Operational traps · Incident log

  • Hooks: Delivery Process Gates

    Blocking checks that enforce phase order, record format and IDs.

    Functions pre-init-spec-check · foundation-readiness-check · gate-sequence-check · status-md-format-check · id-duplicate-guard · store-write-guard · quality-gate-check · estimate-consistency-check · custom-agent-check

  • Hooks: Security & Protection

    Safeguards against destructive actions, unsafe tools and data leakage.

    Functions hitl-destructive-gate · mcp-allowlist-gate · deps-reputation-gate · data-class-router-gate · per-loop-budget-guard · credential-isolation-check · trifecta-guard · tool-cap-check · per-user-audit · memory-poisoning-scan

  • Hooks: Compliance

    Regulatory checks during delivery.

    Functions aiuc-compliance-gate · aiuc-monitor-check · ai-act-gate · gdpr-ropa-check · compliance-evidence · pre-cutover-linter

  • Hooks: Code and Design Quality

    Automatic quality checks after changes.

    Functions post-edit-format · post-edit-typecheck · check-console-log · anti-stub-scan · design-consistency-check · quality-gate-design · seo-geo-lint · trap-scope-check · skill-validation-gate · agent-registry-check

  • Hooks: Telemetry, Costs & Traceability

    Capture costs, times, traces and skill usage for metrics and BeeLens.

    Functions cost-tracker · session-time-tracker · run-metrics-close · subagent-trace-tracker · skill-invocation-logger · context-budget-check · metrics-enforcement-check · trace-completeness-check · pr-created-log · scorecard-on-gate-change · post-delivery-retrospective · lens-step

  • Hooks: Memory & Resumption

    Preserve context across compactions and sessions.

    Functions auto-memory-hook · memory-flush-precompact · compaction-anchor-write · active-task-state

  • Release Gate Scripts & Insights Writers

    Deterministic scripts for the release gate and Insights data (run via commands or danic_script).

    Functions release-gate-check.sh · accessibility-gate-check · advg-dissent-gate · deps-gate · mutation-gate · slop-gate · code-quality-ratchet · review-panel · deps-inventory / deps-license / deps-match · emit-test-coverage / emit-code-quality / emit-mutation-score · emit-ticket-board / emit-metrics-overview / emit-review-metrics · collect-all.js

  • Hooks: Planner/Executor Mode

    Control aids for the cost-saving planner/executor mode.

    Functions planner-executor-trigger · planner-executor-drift · planner-executor-cost-guard

Further OS Building Blocks

Project storage, memory, context, brand, free-text control, headless operation and knowledge base.

  • Delivery Record (Project Artefacts)

    All commands write to a fixed, machine-readable store in the customer repo: feature index, feature folders with status and gates, metrics, quality data and documentation. It is the project's memory and the data source for the BeeKeep3r IDE dashboards.

    Functions feature-index.md · id-registry.md · Feature folders · metrics/ · quality/ & tickets/ · delivery/docs/ · feature-spec/

  • Project Memory (memory.md) & Session Logs

    memory.md holds a project's binding “Active Decisions”; they are loaded at every session start. /daily starts a session, /tldr closes it and updates memory, learnings and feature status.

    Functions Active decisions · Compaction protection · Memory paging · Poisoning scan

  • Operator Context (ABOUT-ME)

    One page about the operator (role, decision rights, language, non-goals, way of working) that agents load at the start of a session; templates, lessons and incidents are loaded only when needed.

    Functions ABOUT-ME.md · TEMPLATES/, lessons/, incidents/, OUTPUTS/

  • Brand Pack (Design Tokens & Constraints)

    Canonical store for a customer's brand specifications: DTCG tokens, constraints, assets, components and Figma references. Customer brand packs are project-owned and survive OS updates.

    Functions tokens/ · constraints.yaml · CLAUDE.md (source priority) · design-debt-log.md · assets/, components/, figma/

  • NL Orchestration (Free-Text Control)

    Free text is mapped automatically to existing commands: the OS detects the project state from delivery/management/ and responds with a [State: …] line and the matching command.

    Functions Intent registry · Workflow templates · State detection

  • Customer-Owned Agents in Delivery Routing

    Customers can register their own agents, which the orchestrators deploy in defined phases; unregistered agents are not routed and may not own gates.

    Functions Registration · Routing rules · Check hook

  • Headless Mode (Autonomous Async Delivery)

    A session controller (Headless Conductor) delivers a validated spec autonomously without asking questions and returns a final report; merge and deployment remain with the human.

    Functions Preflight · DAG & adapters · Budgets · Checkpoints · Limits

  • Mod “beehive-onboarding” (Guided Onboarding)

    Claude Code mod that guides customers step by step from an empty or existing repo to the first delivered feature; detects progress automatically from the delivery artefacts and never starts a command itself.

    Functions Band above the prompt · Greenfield path · Brownfield path · Progress detection · Per project

  • project_nexus: Compliance & Security Scorecards

    Shipped Python package that calculates compliance and security scorecards with Git SHA history and provides further deterministic analyses (legacy extraction, costs, retro).

    Functions compliance · security · extractor · metrics & pricing · quality & retro

  • Knowledge Base (“Agentic Brain”)

    German-language knowledge base of the OS, which agents draw on following a four-stage loading strategy (max. 5–7 mandatory documents per task).

    Functions 00-overview · 01-principles · 02-governance · 03-artifacts · 04-process · 05-roles · 06-templates · 07-learnings · 08-coding-rules / 08-standards · 09-playbooks · 10-architecture · 11-System-Wide-NFR-Framework · 12-skill-evolution-layer · 13-Tools · 14-LocalLLM-danic-onprem · 16-Design-System · 17-Legacy-Mode · 18-software-standards · 19-spec-driven-development · 21-headless-mode

Next step

Turn AI potential into a reliable development process.

Together we assess how DANIC fits your team, your infrastructure and your requirements.

30 minutes · Personal product demo · Google Meet