| name | review-plan |
| description | Review a plan/task/context artifact set before execution, reporting scope gaps, sequencing problems, weak verification, and missing docs/tests/memory updates. |
review-plan
This is the dedicated pre-execution plan review skill.
Use it when the target is a workflow plan, matching task list, and context file,
not source code or a diff.
Defaults
- Default target is the latest valid plan/task/context artifact set under
.agent-layer/tmp/.
- Produce a report only. Do not edit the plan, task, or context artifacts in this workflow.
- If there is no valid plan/task pair, ask for explicit paths or regenerate the artifacts first.
- If the context file is missing, note its absence as a finding.
Required artifact
Write the report to:
.agent-layer/tmp/review-plan.<run-id>.report.md
Use run-id = YYYYMMDD-HHMMSS-<short-rand>.
Create the file with touch before writing.
Artifact discovery
Use the standard artifact naming rule under .agent-layer/tmp/:
<workflow>.<run-id>.plan.md
<workflow>.<run-id>.task.md
<workflow>.<run-id>.context.md
Discovery rules:
- List
.agent-layer/tmp/*.plan.md, .agent-layer/tmp/*.task.md, and .agent-layer/tmp/*.context.md.
- Keep only files that match the standard naming rule and valid
run-id shape.
- Build candidate sets when both
.plan.md and .task.md exist for the exact same <workflow> and <run-id>. A matching .context.md is expected but not required.
- Select the set with the latest
run-id in lexicographic order.
- If the intended set is not the latest valid set, require explicit paths.
Fallback:
- If no valid plan/task pair exists, ask the user for explicit paths or regenerate the plan first.
Multi-agent pattern
Recommended roles:
Plan reader: extracts the stated objective, scope, risks, and exit criteria.
Risk reviewer: looks for missing sequencing, dependencies, and non-goals.
Verification reviewer: stress-tests the test/doc/memory/update expectations.
Reporter: writes the findings report and recommendation.
Global constraints
- Keep the review tied to what the plan actually says, not what you wish it said.
- Produce findings with concrete evidence and exact file references.
- Do not widen this into a code audit. Use the
review-scope skill for code, diffs, or repo slices.
- If the plan is ambiguous, say so explicitly instead of guessing intent.
Human checkpoints
- Required: ask when no valid plan/task pair exists.
- Required: ask when the user intends a non-latest artifact set and explicit paths are needed.
- Optional: ask only when the plan wording is so ambiguous that the review target itself is unclear.
- When a checkpoint involves a genuine tradeoff between substantive alternatives, present at least two options with brief pros and cons, state which you recommend and why, and let the human decide.
- Stay autonomous for normal critique and report writing.
Review workflow
Phase 1: Extract the contract (Plan reader)
From the plan, task, and context artifacts, extract:
- objective
- in-scope items
- explicit non-goals
- sequencing and dependencies
- promised tests or verification
- promised docs or memory updates
- exit criteria
- key files and entry point (from context file)
Phase 2: Critique the plan structure (Risk reviewer)
Check for:
- missing requirements or non-goals
- hidden large refactors
- dependencies ordered after dependents
- risky assumptions presented as settled
- roadmap, issue, or decision constraints that were missed
- context file gaps: missing key files, stale paths, files listed that do not exist, missing entry point
Phase 3: Critique the verification and completion bar (Verification reviewer)
Check for:
- weak or missing verification commands
- missing test work for risky changes
- missing docs or memory updates
- exit criteria that are subjective or not actually testable
- task list items that are too large or vague to execute safely
Phase 4: Record only actionable findings (Reporter)
Each finding must include:
Title
Severity: Critical | High | Medium | Low
Location: exact artifact path and section
Why it matters
Evidence
Recommendation
Required report structure
The report must contain:
# Plan Review Summary
- plan path
- task path
- context path (or note if absent)
- short outcome summary
## Findings
- findings first, ordered by severity
## Open Questions
- only unresolved items that block confidence
## Strengths
- short list of what the plan does well
## Recommendation
approve
approve-with-changes
revise
Guardrails
- Do not report vague “needs more detail” complaints without naming what is missing.
- Do not invent implementation problems that are not implied by the plan.
- Do not collapse multiple plan problems into one oversized finding.
- If the task list or context file is missing but the plan exists, call that out explicitly instead of pretending the artifact set is complete.
Definition of done
- The report exists at
.agent-layer/tmp/review-plan.<run-id>.report.md with every required section (Summary, Findings, Open Questions, Strengths, Recommendation).
- Every finding names its artifact path + section, severity, evidence, and specific recommendation — no vague "needs more detail" entries.
- The report ends with exactly one recommendation:
approve, approve-with-changes, or revise.
- Plan, task, and context artifacts were not modified by this run.
Final handoff
After writing the report:
- Echo the report path.
- Summarize the top findings in chat.
- State the recommendation clearly.