permit0-ai/permit0/reviewing-plan-docs
Claude Code and agent skill by permit0-ai/permit0
- community
Checking that the SKILL.md file is still in the repository.
Review planning documents by reading the plan docs and doing a thorough codebase walkthrough, then writing detailed review docs. Use when the user asks to review, audit, critique, or validate a plan, design doc, or spec in the permit0 repository.
Install
npx vskill@latest install permit0-ai/permit0/reviewing-plan-docsSKILL.md
Reviewing Plan Docs for permit0
When to Use
Use this skill when the user asks you to review, audit, critique, or validate planning documents under docs/plans/<feature>/. The output is a set of review docs written to docs/plan-reviews/<feature>/.
Workflow
Phase 1: Read the Plan
Read every file in docs/plans/<feature>/ in numbered order. For each doc, note:
- What it claims (the design decisions, schemas, code changes).
- What assumptions it makes.
- What it explicitly defers or omits.
Phase 2: Codebase Walkthrough
Do a thorough read of the actual source code that the plan references. This is not optional -- you must verify claims against real code.
For integration plans, always read:
- The reference implementation the plan says it is based on. For example,
if the plan says "replicate the Claude Code integration," read crates/permit0-cli/src/cmd/hook.rs end to end.
- The crate(s) the plan says it will modify. Read every file in those
crates, not just the ones the plan mentions.
- The crates the plan says are NOT modified. Spot-check at least the
public API surface to verify the plan's claim that no changes are needed.
- Packs and risk rules if the feature touches normalization or scoring
(read packs/permit0/email/ as the reference pack).
- The daemon/serve code if the feature involves
--remotemode
(crates/permit0-cli/src/cmd/serve.rs).
- Tests -- read existing test files to understand current coverage and
what the plan's new tests need to complement.
Phase 3: Write the Review
Write review docs to docs/plan-reviews/<feature>/. One review file per plan doc, plus a summary.
Output Structure
Skip review files for plan docs that don't exist (e.g. if there is no 01-protocol.md, skip 02-review-protocol.md).
Review Doc Format
Every review doc must follow this structure:
Summary Doc Format (00-summary.md)
Distinguishing Your Review
Multiple agents may review the same plan. To make your review distinguishable:
- Always include the reviewer line: `Reviewer: Cursor Agent
(session <unique identifier>)` -- use the current timestamp or conversation ID.
- Cite specific file paths and line numbers from the codebase when
verifying or disputing claims. Other reviewers may make the same point but your evidence trail will differ.
- Include a "Verified Claims" section -- this is your proof-of-work
showing which parts of the codebase you actually read.
What Makes a Good Review
DO
- Verify every code snippet in the plan against the actual source.
Plans often show idealized code that doesn't match current function signatures or module structure.
- Check for missing error handling. Plans often describe the happy path
and skip what happens on failure.
- Validate the "files NOT changed" list. If the plan claims the engine
is untouched, confirm no engine changes are actually needed.
- Look for wire-format mismatches. If the plan describes a JSON schema
for communication between two components, verify both sides agree. This is where bugs hide (e.g. one side sends "human", the other expects "humanintheloop").
- Check that test coverage matches the claimed behavior. If the plan
says "HITL maps to deny," there should be a test for that.
- Flag implicit assumptions. If the plan assumes network access in a
sandboxed environment, call it out.
DON'T
- Don't rubber-stamp. If you only found nits, you probably didn't read
deeply enough.
- Don't rewrite the plan. Findings should be specific and actionable, not
alternative designs.
- Don't skip the codebase walkthrough. A review based only on reading the
plan docs is not a review -- it's proofreading.
Severity Definitions
| Severity | Meaning | |
|---|---|---|
| Critical | The plan is wrong in a way that would cause a security hole, data loss, or silent governance bypass if implemented as written. | |
| Major | The plan has a significant gap or incorrect assumption that would cause bugs or user confusion, but not a security issue. | |
| Minor | The plan is imprecise or incomplete in a way that would slow down implementation but not cause defects. | |
| Nit | Style, naming, or documentation improvement. |
Security
- Security scanPassed, 100 of 100Pattern checks on the skill's files.
- Model reviewNot runA model review is scheduled for the next rescan.
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/reviewing-plan-docs/SKILL.md
Popularity
- Repository stars
- 202
- Repository forks
- 21
- 7-day trend score
- 27
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/permit0-ai/permit0/reviewing-plan-docs)