council-workflow

Orchestrates multi-agent PRD and ADR creation with resumable document-driven state tracking.

Updated Mar 4, 2026
One-click install
npx skills add https://github.com/chisuhua/home --skill council-workflow-chisuhua
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: council-workflow
Source: https://github.com/chisuhua/home/tree/main/.openclaw/skills/council-workflow
Command: npx skills add https://github.com/chisuhua/home --skill council-workflow-chisuhua

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Coordinating research, product, architecture, and review discussions for PRD/ADR creation often loses progress when sessions are interrupted, and decision points get confirmed without real discussion. This Skill runs a document-driven state machine that tracks every step in a todo file, so any interruption can be resumed from the exact breakpoint. ## Core Features & Use Cases - Multi-Agent Council Workflow: Spawns researcher, PM, architect, reviewer, and librarian subagents in sequence (with PM and architect running in parallel), each producing verified Markdown outputs. - Breakpoint Recovery: Uses a .council-todo.md file as the single source of truth, so restarting the skill resumes from the current step instead of starting over. - Active Decision Discussion: Instead of sending a bare list of decision points, the director agent analyzes each point, gives recommendations with reasoning, and discusses with the user before confirmation. - Use Case: A team needs a PRD and ADR for a new caching feature. The skill researches options, gathers PM and architect perspectives, has a reviewer reconcile disagreements, discusses trade-offs with the user, then generates both documents. ## Quick Start Start the council workflow for the topic "user session caching" in my project and generate the PRD and ADR.

Frequently Asked Questions about council-workflow

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a PRD and ADR with multiple AI agents?▼

Run the council workflow with a topic, and it spawns researcher, PM, architect, reviewer, and librarian subagents in sequence. Each agent writes a verified Markdown file, and the final step generates the PRD and ADR from user-confirmed decisions.

How does the council workflow recover after an interruption?▼

It reads the .council-todo.md state file on startup and checks each step's status. Steps marked done are skipped, in-progress steps are validated against their output files, and failed steps are reported to the user for retry or termination.

Can PM and architect discussions run in parallel?▼

Yes, Step 3 (PM) and Step 4 (architect) can be spawned simultaneously since both only depend on the research output. The workflow waits until both return and their files are verified before starting the reviewer step.

What happens if a subagent output file is missing or empty?▼

The workflow respawns the subagent once. If the file is still missing or empty after the retry, the step is marked failed in the todo file and the error is reported to the user.

What templates does the workflow use for PRD and ADR documents?▼

The references directory provides a PRD template with MoSCoW prioritization, user stories, and acceptance criteria, plus an ADR template with alternative options, decision rationale, risks, and technical debt sections. A review checklist validates cross-consistency between the two documents.