review-code-plan-satisfaction
Reviews implementation fidelity against project plans and amendments to ensure accuracy.
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.zipInstalls 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.Key capabilities
- →Check implementation fidelity against PLAN.md and AMENDMENTS.md
- →Identify silent deviations and missing approved amendments
- →Verify validation facts and proof of claimed changes
- →Flag unauthorized backward-compatibility work
- →Write or update PLAN_SATISFACTION_REVIEW.md with findings
- →Create or update CODE_REVIEW.md with compliance verdict
How it works
The skill performs a multi-step review process, comparing code changes against documented plans and amendments, validating claimed behaviors, and generating structured findings and compliance reports.
Inputs & outputs
When to use review-code-plan-satisfaction
- →Audit code fidelity
- →Verify project amendments
- →Review implementation logic
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.mdwhen reviewing an initiative child packet../../references/communication-mode.md- relevant tracked
AGENTS.mdchain - applicable
AGENTS.override.mdonly if local execution constraints matter - current task packet:
PLAN.md,AMENDMENTS.md,INDEX.md, optionalCODE_REVIEW.mdif 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_NOTEmarkers 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
- 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.
- 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.
- Compare the implemented change against
- 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.
- 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.
- 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.
- 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.
- Emit compact
- Create or update
CODE_REVIEW.mdwith 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, ornon_compliant- unresolved
PS# - latest validation summary
Phase 2 handoff:concise note for maintainability review consumption- key dispositions already decided in
AMENDMENTS.md
- Update
INDEX.mdif 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 InitiativeandImported Context - declared dependencies and contract surface still match the implementation
- optional
CONTRACT_DECISION.mdwas followed when present PLAN.mdwas authored, not left asStatus: 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.
When not to use it
- →When performing a deep maintainability, evolvability, or anti-slop review
- →When writing tracked AGENTS.md or AGENTS.override.md
- →When the task is to bless claimed validation that was not actually run
Limitations
- →Not for deep maintainability/evolvability/anti-slop review
- →Does not bless claimed validation that was not actually run
- →Does not confuse desired outcome with plan compliance
How it compares
This skill provides a structured, findings-first review focused on plan fidelity and validation proof, ensuring compliance and correctness before broader maintainability checks.
Compared to similar skills
review-code-plan-satisfaction side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| review-code-plan-satisfaction (this skill) | 0 | 3mo | No flags | Advanced |
| python-testing-patterns | 77 | 3mo | Review | Intermediate |
| dependency-upgrade | 26 | 5mo | Review | Intermediate |
| test-cases | 57 | 7mo | No flags | Beginner |
Try saying
Example prompts that trigger this skill in your AI assistant.
You might also like
python-testing-patterns
wshobson
Implement comprehensive testing strategies with pytest, fixtures, mocking, and test-driven development. Use when writing Python tests, setting up test suites, or implementing testing best practices.
dependency-upgrade
wshobson
Manage major dependency version upgrades with compatibility analysis, staged rollout, and comprehensive testing. Use when upgrading framework versions, updating major dependencies, or managing breaking changes in libraries.
test-cases
cexll
This skill should be used when generating comprehensive test cases from PRD documents or user requirements. Triggers when users request test case generation, QA planning, test scenario creation, or need structured test documentation. Produces detailed test cases covering functional, edge case, error handling, and state transition scenarios.
reviewing-code
CaptainCrouton89
Systematically evaluate code changes for security, correctness, performance, and spec alignment. Use when reviewing PRs, assessing code quality, or verifying implementation against requirements.
wcag-audit-patterns
wshobson
Conduct WCAG 2.2 accessibility audits with automated testing, manual verification, and remediation guidance. Use when auditing websites for accessibility, fixing WCAG violations, or implementing accessible design patterns.
code-coverage-with-gcov
gadievron
Add gcov code coverage instrumentation to C/C++ projects