| name | plan-review-invoke |
| description | Use when reviewing an implementation plan in a fresh session, after the plan was written in a prior session, and you want an adversarial pass (checklist + elegance gate + a self-contained handoff plan) without copy-pasting a reviewer prompt. Spawns the shipped plan-reviewer agent on the plan file. Accepts the plan path as an argument. NOT for: same-session plan reviews (invoke the plan-reviewer agent directly with the plan already in context); reviewing implementations or diffs that already shipped (use a code-review or session-critique flow); or the blind frame-breaking pass (use elegance-review).
|
| argument-hint | <path-to-plan> |
plan-review-invoke
Runs the plan-reviewer agent (agents/plan-reviewer.md) against a plan file
that was authored in an earlier session. The agent carries the full reviewer
contract itself, the read-before-critique ground rule, the six-point checklist,
the elegance gate, and the handoff-plan output, so this skill only has to resolve
the path and pass the plan in with a cross-session note.
When to fire
User types /plan-review-invoke <path-to-plan> in a fresh session, for example:
/plan-review-invoke .claude/plans/new-feature.md
/plan-review-invoke docs/plans/refactor-auth.md
/plan-review-invoke (no path, ask which plan)
Skip if:
- Same-session review: the planning session already has full context, so invoke
the plan-reviewer agent directly via the Agent tool with a focused prompt.
- The target is an implementation diff, not a plan: use a code-review or
session-critique flow instead.
- The target is a spec or contract document rather than a plan.
Process
-
Resolve the plan path. If the user supplied a path, confirm the file
exists with Read. If none was supplied, ask "Which plan should I review? Paste
the path." and confirm before proceeding.
-
Invoke the agent. Spawn the reviewer via
Agent(subagent_type: "plan-reviewer", ...). The prompt is:
<plan_to_review>
{{full contents of the plan file at the user-supplied path}}
</plan_to_review>
This plan was created in a PRIOR session, not this one. Treat the
<plan_to_review> block as the authoritative plan content and the repository
as authoritative for what is true. Follow your full reviewer contract:
read before critiquing, work the checklist, run the elegance gate, and emit
a self-contained handoff plan plus your verdict.
-
Surface the verdict. The agent returns APPROVE / APPROVE WITH CHANGES /
REJECT with findings and a handoff-ready revised plan. Present the verdict and
key findings to the user. Do NOT auto-apply plan changes; the user decides
whether to revise, and the calling session writes any revised plan to disk.
Why a skill and not just the agent
Invoking the agent by hand every time means restating the cross-session framing
and remembering to pass the plan file's contents rather than a path the agent
cannot see. This skill collapses that to one command and guarantees the plan text
is read in fresh at invocation time.
See also
agents/plan-reviewer.md: the reviewer's behavior (checklist, elegance gate,
handoff contract, verdict).
agents/elegance-reviewer.md and the elegance-review skill: a separate
blind frame-breaking pass. For high-stakes plans, run it after this one as an
independent review, and never pass it the author's reasoning.
workflows/plan-review-fanout.js: a parallel, grounded upgrade of this skill
that fans the same checklist across several lenses.