How this skill is triggered — by the user, by Claude, or both
Slash command
/plan-review:plan-with-reviewThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Two modes:
Two modes:
/plan-with-review) — user-gated. Show the staff engineer's review and wait for the user to decide what to apply./plan-with-review auto) — incorporate the review automatically with no user gate. Use only when the user includes the literal token auto in the invocation.staff-engineer subagent (see Prompt-framing rules below).If the staff engineer's review contains no items in any of the four sections, skip steps 3–5 and present the original plan as final.
When auto appears as an argument, run steps 1–2, then automatically incorporate every actionable item from the review and present the revised plan. Do not ask the user anything. This matches the prior behavior of this skill.
When delegating the plan to the staff-engineer subagent:
The reviewer should encounter the plan as an anonymous document, not as something you are asking them to approve.
Show the staff engineer's verbatim output (all four sections: Critical Issues, Significant Concerns, Questions, Minor Suggestions) exactly as returned. Then a horizontal rule (---). Then a flat numbered action list covering every concrete item across all four sections, one per line:
1. [Critical] <one-sentence summary>
2. [Significant] <one-sentence summary>
3. [Question] answer: <the question, condensed>
4. [Minor] <one-sentence summary>
End with this exact prompt to the user:
Reply with which items to apply (e.g. "apply 1, 3, 5; skip 2; for 4 do X instead"), or
all/none.
Accept any of:
all — apply every item.none — apply nothing; revised plan equals the original draft. State explicitly that no changes were applied.1, 3, 5 or 1-3) — apply only those.apply 1 and 3, skip 2, for 4 do X instead) — apply the listed items, honoring overrides.Only ask a follow-up clarifying question if the reply is genuinely ambiguous. Otherwise proceed directly to producing the revised plan.
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 market-math/mm-claude-plugin --plugin plan-review