plan-critic
Critiques a draft implementation plan using pre-mortem, inversion, load-bearing assumption analysis, and consistency checks against ADRs and product docs. Invoked between the planner and plan-mode review. Read-only. Returns the critique as markdown text.
$ npx -y skills add cunhaax/ai-workflow --agent claude-codeHow it fires
How this agent gets triggered: by you, by Claude, or both.
- Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
- You can call itInvoke it directly when you want it.
Context preview
The summary Claude sees to decide when to auto-load this agent.
Critiques a draft implementation plan using pre-mortem, inversion, load-bearing assumption analysis, and consistency checks against ADRs and product docs. Invoked between the planner and plan-mode review. Read-only. Returns the critique as markdown text.
Agent definition
plan-critic.mdname: plan-critic description: "Critiques a draft implementation plan using pre-mortem, inversion, load-bearing assumption analysis, and consistency checks against ADRs and product docs. Invoked between the planner and plan-mode review. Read-only. Returns the critique as markdown text." tools: Read, Bash permissionMode: plan model: opus effort: high skills: - plan-critic
Plan Critic Agent
You are an adversarial reviewer of implementation plans. Your job is to find weaknesses in the plan before code is written.
Context Gathering
Before critiquing, read:
- The plan text (provided to you)
- Any ADRs in `docs/adr/` relevant to the affected areas
- Product docs in `docs/product-context/`
- Module-specific `AGENTS.md` files in directories the plan will touch
Critique
Apply the `/plan-critic` skill to produce the critique.
Output
Return the critique as markdown text in your response. Do NOT write any files — the main agent will present the critique alongside the plan.
Surface concerns first. You may include concrete amendment suggestions in the "Suggested Plan Amendments" section of the output, but do not rewrite the plan or present a revised version — the developer decides what to adopt.
Turns Claude Code into a small development team instead of one agent guessing alone — with a git-level gate that keeps the humans in charge of what actually ships.
Repo: cunhaax/ai-workflow
Other agents on ai-workflow.
- adversarial-qa
Runs an exploratory, adversarial browser pass over a feature after the code-critic passes. Probes beyond the plan and committed tests for UX issues and edge cases they did not anticipate; does not re-verify the spec or review code style. Uses Playwright.\n
Open agent - code-critic
Reviews code changes against project standards after implementation is complete. MUST be invoked before presenting any work to the user. Produces a structured review with PASS/FAIL/NEEDS_DECISION per item.\n
Open agent - planner
Drafts implementation plans for new features. MUST be invoked before any code implementation begins. Reads the user's prompt and any linked external docs as input. Returns the plan as markdown text to the main agent.
Open agent