AI-DLC development harness collection. Brings the AI-Driven Development Lifecycle (Idea → Proposal → Task → Execute → Verify → Done) to popular project management platforms via Claude Code plugins.
Part of the Chorus AI-DLC ecosystem — an open collaboration platform for AI Agents and humans.
AI-DLC is a structured workflow where AI agents and humans collaborate through defined roles:
- PM Agent — discovers ideas, runs structured elaboration, creates proposals with PRD and task breakdowns
- Developer Agent — claims tasks, implements code, self-checks acceptance criteria with evidence
- Admin Agent — reviews proposals, verifies tasks against AC, manages project governance
The workflow follows: Idea → Elaboration → Proposal → Approval → Execute → Verify → Done
Each stage has dedicated skills with Linear-specific guidance, and the plugin automates session lifecycle, context injection, and observability through Claude Code hooks.
| Plugin | Platform | Version | Status |
|---|---|---|---|
| linear | Linear | v0.4.0 | Active |
- 6 modular skills — core, idea, proposal, develop, quick-dev, review
- Independent review agents —
proposal-reviewerandtask-reviewerauto-suggested via PostToolUse hooks after submission - Structured elaboration — multi-round Q&A via
AskUserQuestionwith categories and depth levels - Proposal label state machine — draft → submitted → approved/rejected via label stacking
- Structured acceptance criteria —
AC-{n}:format with dev self-check and admin verification - Quick-Dev fast track — skip Idea→Proposal for small tasks, with admin self-verify
- Agent Teams integration — wave-based parallel execution with DAG-aware task dispatch
- Automated hooks — session lifecycle, PostToolUse review triggers, unblocked task discovery
- Configurable via userConfig —
enableProposalReviewer/enableTaskReviewertoggles in plugin settings - Client-side sessions — no server dependency, full observability via Linear Issue Comments
# 1. Add the marketplace
claude /plugin marketplace add Chorus-AIDLC/Development-Agent-Harness
# 2. Install the Linear plugin
claude /plugin install linear@development-agent-harness
# 3. Set your Linear API key
export LINEAR_API_KEY="lin_api_xxx"harness:* labels are auto-created on first session start — no manual bootstrap needed.
- Start a session: Labels are auto-created on first launch
- Check your role and use the appropriate skill:
- PM Agent →
/ideathen/proposal - Developer Agent →
/develop - Admin Agent →
/review - Quick tasks →
/quick-dev
- PM Agent →
- For Agent Teams: Team Lead spawns sub-agents with
linear:issue:<identifier>— the plugin auto-injects session context
.claude-plugin/ # Marketplace manifest
plugins/
linear/ # Linear platform plugin
.claude-plugin/ # Plugin manifest + userConfig
.mcp.json # MCP server config
hooks/ # Claude Code hook routing (incl. PostToolUse)
bin/ # Hook scripts + CLI tools
agents/ # Independent review agents
proposal-reviewer.md # Read-only proposal QA (VERDICT: PASS/FAIL/PARTIAL)
task-reviewer.md # Read-only task verification (runs tests, checks AC)
skills/
linear-harness/ # Core skill — overview, shared tools, setup
idea/ # Ideation — elaboration via AskUserQuestion
proposal/ # Planning — PRD, tasks, DAG, label state machine
develop/ # Development — claim, implement, AC self-check
quick-dev/ # Fast track — skip proposal for small tasks
review/ # Governance — approve/reject, verify with AC
This harness brings the Chorus AI-DLC workflow to Linear. Chorus is a full collaboration platform with server-side session management, structured MCP tools, and a web UI. This harness adapts the same concepts for Linear using:
- Linear Issues as the equivalent of Chorus Ideas, Proposals, and Tasks
- Linear Documents as the equivalent of Chorus PRD/Tech Design documents
- Linear Labels (
harness:*) as the workflow state machine - Linear Issue Relations (blocking/blocked-by) as the task dependency DAG
- Linear Issue Comments as the observability trail (replacing Chorus server-side sessions)
- Client-side state files (
.linear-harness/) for local session management
The design philosophy is the same — "Reversed Conversation" where AI proposes and humans verify — adapted to work natively with Linear's data model.
MIT