agentskills.codes
RE

review-code-plan-satisfaction

Review implementation fidelity against `PLAN.md` and `AMENDMENTS.md`; emit `PS#` findings and validation-fact gaps into the task packet.

Install

mkdir -p .claude/skills/review-code-plan-satisfaction && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/16957" && unzip -o skill.zip -d .claude/skills/review-code-plan-satisfaction && rm skill.zip

Installs to .claude/skills/review-code-plan-satisfaction

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Review implementation fidelity against `PLAN.md` and `AMENDMENTS.md`; emit `PS#` findings and validation-fact gaps into the task packet.
136 charsno explicit “when” trigger

About this skill

review-code-plan-satisfaction

Phase 1 code review. Check whether the implemented change matches the approved plan and whether the claimed behavior is actually proved.

Keep this pass narrow, concrete, and findings-first.

Read first

  • ../references/task-packet-contract.md
  • ../references/persistent-artifacts-contract.md
  • ../references/review-phases-contract.md
  • ../references/compatibility-policy.md
  • ../references/final-state-authoring-policy.md
  • ../references/initiative-workflow-contract.md when reviewing an initiative child packet
  • ../../references/communication-mode.md
  • relevant tracked AGENTS.md chain
  • applicable AGENTS.override.md only if local execution constraints matter
  • current task packet: PLAN.md, AMENDMENTS.md, INDEX.md, optional CODE_REVIEW.md if already present

Core scope

Check:

  • fidelity to PLAN.md
  • fidelity to AMENDMENTS.md
  • silent deviations
  • missing approved amendments
  • plan sequencing/design drift that should have triggered amendment
  • validation facts: what was actually run, and whether it proves the claimed change
  • whether REVIEW_NOTE markers were used only for non-blocking better alternatives
  • unauthorized backward-compatibility work or unresolved compatibility decisions at real external boundaries
  • canonical owner artifacts that still diverge from the plan because old/new vocabulary or status semantics were both kept alive

Not core scope:

  • deep maintainability/evolvability/anti-slop review

Core workflow

  1. Establish review scope.
    • Identify the exact implementation commit under review, the exact diff from that commit, touched files, tests added or changed, and any untracked files that are intentionally part of the implementation.
    • Prefer reviewing the real patch over summaries from memory.
    • If the lead did not provide the commit hash and the workspace is dirty, stop and ask which commit is the review target instead of guessing from a floating workspace.
    • If the workspace has drift beyond the named commit, review the named commit and call out the drift explicitly so the lead can re-anchor the review.
  2. Run a plan-fidelity pass.
    • Compare the implemented change against PLAN.md, not just against the desired outcome.
    • Flag silent deviations, missing amendments, and design changes made without approval.
    • Treat unapproved backward-compatibility code as non-compliant.
    • If the change touches a real external boundary and the packet never resolved whether backward compatibility was required, fail the handoff and require explicit human clarification.
  3. Run a final-state fidelity pass.
    • Flag implementations that keep superseded contract vocabulary, owner semantics, or status models alive inside canonical artifacts when the approved plan chose one final model.
    • Flag code/docs/contracts that read like migration residue instead of the approved target state.
  4. Run focused validation.
    • Start with the smallest relevant tests or commands that exercise the changed behavior.
    • Expand only as needed for confidence.
    • Record what was actually run.
  5. Run a test-adequacy pass.
    • Treat a weak test as a finding when it claims to validate the change but would still pass if the bug remained.
    • Prefer one strong targeted test over a broad noisy suite.
  6. Write or update PLAN_SATISFACTION_REVIEW.md.
    • Emit compact PS# findings ordered by severity.
    • Keep the explanation concrete and tied to the code path or missing proof.
  7. Create or update CODE_REVIEW.md with the fixed rollup fields this phase owns.
    • If the file is missing, initialize it with the full canonical schema before handing off to phase 2.
    • Compliance status: compliant, compliant_with_amendments, or non_compliant
    • unresolved PS#
    • latest validation summary
    • Phase 2 handoff: concise note for maintainability review consumption
    • key dispositions already decided in AMENDMENTS.md
  8. Update INDEX.md if review status changes.

Findings bar

Only report a PS# finding when all are true:

  • it is actionable
  • it affects correctness, compliance, or confidence in the claimed behavior
  • you can point to the exact code path, plan step, amendment, missing proof, or unauthorized compatibility layer

How to inspect changes

  • Start from the named implementation commit.
  • Prefer git show --stat <commit>, then targeted diff reads, then final-code reads.
  • Use workspace diff only to detect drift beyond the named review commit, not to silently move the review target.
  • If behavior depends on runtime state, inspect both the writer and the reader.
  • If behavior depends on persistence or recovery, inspect creation, update, load, and restart paths together.
  • When reviewing tests, ask whether the assertion would fail if the bug still existed.

Requester-response rule

Always report the review result back to the requesting side.

  • If the lead/orchestrator requested the review, reply back to the lead/orchestrator.
  • If a human directly requested the review in this pane, reply back in this pane.
  • Local pane output without an explicit reply to the requester does not count as review completion.

Output shape

Default output should include:

  • findings first, highest severity first
  • exact PS# findings
  • concise validation notes
  • explicit compliance verdict
  • residual validation gaps if there are no findings

Do not

  • collapse this into a generic maintainability pass
  • bless claimed validation that was not actually run
  • confuse desired outcome with plan compliance
  • write tracked AGENTS.md
  • write AGENTS.override.md
  • treat unapproved compatibility shims as harmless implementation detail
  • accept canonical code/docs/contracts that visibly preserve both an old and a chosen new model

Initiative child packets

For initiative-created child packets also check:

  • alignment with Parent Initiative and Imported Context
  • declared dependencies and contract surface still match the implementation
  • optional CONTRACT_DECISION.md was followed when present
  • PLAN.md was authored, not left as Status: not_authored

Communication

Honor active caveman mode for user-facing replies per ../../references/communication-mode.md. Keep durable artifacts normal unless the human asks otherwise. Drop caveman for safety/clarity when needed, then resume.

Search skills

Search the agent skills registry