wipnoteplan-critiquelisted

# wipnote Plan Critique Invoke this skill **after** the plan is drafted via `wipnote:plan` (slices laid out, decisions_notes captured) and **before** the human review pass. It runs two critic agents in parallel against the plan YAML and bakes the findings into the affected slices as `critic_revisions`. **Trigger keywords:** critique this plan, review plan, plan critic pass, run critics, design critique, feasibility critique --- ## When to run | Plan shape | Critique? | |---|---| | Trivial single-slice plan | Skip — overhead outweighs value | | Standard plan with 2+ slices | Run | | Complex plan (any slice with `complexity: complex`) | **Required** before human review | --- ## Invocation There is no `wipnote plan critique <plan-id>` subcommand. The agent invokes this skill directly, runs both critic prompts against the plan YAML, and writes findings back into the slice cards. The agent: 1. Loads `.wipnote/plans/<plan-id>.yaml` and reads each slice card. 2. Runs two critic passes itself (no separate CLI dispatch): - **DESIGN CRITIC** — Haiku model. Reads the plan as a design document. Looks for missing slices, scope gaps, unclear acceptance criteria, conflicting goals, undefined lifecycle states, and reviewer ergonomics issues (will a human actually understand and approve this?). - **FEASIBILITY CRITIC** — Sonnet model. Reads the plan as an implementation contract. Verifies cited files exist, line ranges match, existing patterns are reused not reinvented, gates