From movp
Adversarial review for plan files and significant code output. Invoke after writing a plan file, completing a multi-file code change, or when the user asks for a review. Do not invoke for trivial edits, file reads, or git commands.
How this skill is triggered — by the user, by Claude, or both
Slash command
/movp:movp-reviewThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
> Scope: plan files and significant code output
Scope: plan files and significant code output
MCP tools:get_review_status,trigger_review,resolve_review
After writing a plan file or producing a significant code change (new feature, refactor, migration, or multi-file edit), trigger an adversarial review using the MoVP tools. Do not trigger for trivial operations: file reads, git status/log, search queries, single-line fixes, or tool calls with no code output.
Trigger trigger_review automatically when you:
.md planning documents, architecture docs, PRDs)Do not trigger for: reading files, running tests, git commands, grep/search, single-line edits, config-only changes.
1. Call trigger_review(artifact_type="plan_file"|"code_output", content=<artifact>, session_id=<current session>)
→ returns review_id
2. Poll get_review_status(review_id=<id>) until review_status is "completed" or "error"
Always use the review_id from step 1 — do NOT call get_review_status without review_id
when multiple reviews may be in flight (returns most recent tenant review otherwise).
3. Ask the developer: implement fixes, dismiss findings, or accept as-is
4. Call resolve_review(review_id=<id>, action="accept"|"dismiss"|"escalate"|"retry") based on their choice
Format findings as structured output with severity badges. After showing findings, always ask:
Reply with: implement fixes, dismiss (false positive / not applicable / deferred), or accept as-is
| Developer says | Action to call | Notes |
|---|---|---|
| "accept", "looks good", "ship it" | resolve_review(action="accept") | Idempotent — safe to call twice |
| "dismiss", "false positive", "not applicable" | resolve_review(action="dismiss", reason="false_positive"|"not_applicable"|"deferred") | |
| "escalate", "create a ticket" | resolve_review(action="escalate", target="todo") | |
| "retry", "run it again" | resolve_review(action="retry") | Only valid when review_status is "error" — do not call on completed reviews |
Reviews consume LLM budget. Do not trigger multiple reviews in a single session for the same artifact. If trigger_review returns a rate limit error (429), inform the developer and do not retry automatically.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub mostviableproduct/movp-plugins --plugin movp