review
Skill566 starsChanged 14 days agoApache-2.0View on GitHub
High-signal, embedded-aware Pigweed code review skill with strict comment calibration. Dispatches the enabled language reviewer subagents in parallel (one patch each), merges their findings into one calibrated verdict, and falls back to an inline audit for files with no enabled reviewer. Supports both Self-Review Gate mode (pre-upload verification for /crank) and Gerrit CL Reviewer mode (staging calibrated --draft comments via ./gh pr comment). Do not load for a quick LGTM or a single-file look, and not for /respond-style comment handling (see .agents/skills/respond/SKILL.md).
- Commits and pushes
192 other projects use a version of this.
Adopted by
Most recently taken up by
- dzhng/skillstheir own version, first seen Nov 15, 2026
- muxy-app/muxytheir own version, first seen Nov 15, 2026
- macro-inc/macrotheir own version, first seen Nov 15, 2026
- igoroliveirg/claude-copytheir own version, first seen Nov 15, 2026
- yagizdo/quivertheir own version, first seen Nov 15, 2026
- markmhendrickson/neotomatheir own version, first seen Nov 15, 2026
Used with
Projects with it also use close-spec, daily-arxiv and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
dzhngSkillclose-spec
When a spec under specs/<feature>/ is done shipping (a write-spec build, or any planned task that produced a spec), archive it to specs/done/ and rewrite it from a build-plan into a durable rationale — the why, the principles, the invariants — pointing back to the real code for the how. Use when implementation has landed and the plan no longer matches what shipped, or the user says a feature/spec is finished. Pairs with [write-spec](../write-spec/SKILL.md) (the plan this closes) and [review](../review/SKILL.md) (run before closing).
Also in 5 projects that use this
LambenthanSkilldaily-arxiv
每日从 arXiv 拉取新论文,过滤相关性,auto-ingest 高优先级论文,检测 SOTA 更新
Also in 5 projects that use this
dzhngSkilldesign-with-images
Explore UI and visual design with image generation, then iterate the real implementation against the user-selected concept. Use when brainstorming visual directions, offering mockup variants, or implementing an approved generated design; do not stop at the mockups or the first code approximation.
Also in 5 projects that use this
LambenthanSkillexp-eval
实验判决门:Review LLM 独立评判实验结果 → 4 种判决路径 → 自动更新 claims confidence、ideas status、graph edges
Also in 5 projects that use this
- +5 more from Lambenthan
- +3 more from dzhng
Other versions192 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 192
- garrytan/gstack★ 142ksame name
- ansible/ansible★ 71ksame name
- docling-project/docling★ 71ksame name
- traefik/traefik★ 66ksame name
- ClickHouse/ClickHouse★ 51ksame name
- temporalio/temporal★ 24ksame name
- ccch1mneyyy/dsh-Go★ 24ksame name
- tisfeng/Easydict★ 15ksame name
- juicedata/juicefs★ 15ksame name
- elie222/inbox-zero★ 13ksame name
182 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository21 other files it gives its agents.
- AGENTS.mdAGENTS.md
- pw_ghish/AGENTS.mdAGENTS.md
- pw_ide/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- BluetoothSkill
- Code ReviewSkill
- crankSkill
- docsSkill
- freshenSkill
- ghishSkill
- learn-pigweedSkill
- manual-cli-testSkill
- oncallSkill
- pw_async2Skill
- Pigweed Kernel TestingSkill
- C++ Pigweed RPC ServicesSkill
- respondSkill
- run-presubmit-checksSkill
- saleae-analyzerSkill
- Use Pigweed Toolchain ToolsSkill
- Multi-Agent Worktree & Bazel Cache ManagementSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
20 rules. 0 are common, 5 are distinctive, and 15 are very specific to this project.
0
Common
Most of the 156 edited versions we read keep these.
5
Distinctive
This author's own choices. Worth a look.
15
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 52
Every comment or finding MUST pass these calibration filters:
- Distinctive: line 54
Zero Praise / Filler Comments: Never post inline comments that merely say *"Nice refactor!"*, *"Looks good"*, or restate what the code does.
- Very specific: line 56
Verify Before Accusing: Before claiming a symbol is undefined, a header is missing, or a lock is unlocked, verify with code_search, …
- Distinctive: line 59
Mandatory Severity Prefix: Every inline finding must begin with one of:
- Very specific: line 76
Respect Formatter Authority: Never leave inline comments about whitespace, line wrapping, or include sorting that ./pw format handles …
- Very specific: line 81
Cite the Rule: When a finding rests on a Pigweed rule, link it as a pigweed.dev URL, e.g.
- Very specific: line 102
If the CL links a Buganizer issue (Bug: b/<id> or Fixed: b/<id>), inspect it via ./gh issue view <id> to verify the change actually solves …
- Very specific: line 105
Hard stops: If no target resolves (no HEAD ahead of origin/main and no CL given), or the CL is MERGED or ABANDONED (./gh pr view <cl> …
- Distinctive: line 137
Wait for every reviewer to finish before merging.
- Very specific: line 162
ISR & Thread Safety (
pw_sync): Check InterruptSpinLock vs. - Very specific: line 173
pw_async2(if applicable): Verify Pend(Context&) implementations properly register a Waker (PW_ASYNC_STORE_WAKER) whenever returning … - Very specific: line 176
Rust /
pw_kernel(if applicable): Verify no_std compliance, minimal unsafe blocks with explicit // SAFETY: justifications, and panic-free … - Very specific: line 184
If a source file (.cc), public header (.h), or dependency was added or removed in BUILD.bazel, verify whether the module also maintains …
- Very specific: line 190
Every bugfix or behavior change requires a focused pw_unit_test (*_test.cc) test case.
- Very specific: line 201
Commit message must follow <module>: <Imperative subject <=72 chars>, explain the *why*, include Bug: b/<id> or Fixed: b/<id>, and preserve …
- Very specific: line 205
If the change touches .agents/, AGENTS.md, GEMINI.md, or CLAUDE.md: read docs/sphinx/style/ai_artifacts.rst in full at review time and …
- Very specific: line 215
When running in Mode A (or Mode B from a worktree allocated with ./gh wt use review-<cl> --cl <cl> --json.
- Distinctive: line 218
Run hermetic tests for modified modules:
- Very specific: line 241
Check existing comments (./gh pr view <cl> --comments) so you do not duplicate feedback already left by another reviewer.
- Very specific: line 247
Put out-of-diff findings, if any, in one change-level draft (./gh pr comment <cl> -m "<summary>" --draft).
“Common” is measured: kept by at least 6 in 10 of the 156 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Pigweed Calibrated Code Review (/review)
/review performs a rigorous, high-signal engineering review tailored to
Pigweed's embedded C++, Rust (pw_kernel), Python, and dual-build (Bazel +
GN) architecture.
Unlike generic review prompts that eitherrubber-stamp changes or flood CLs with
low-value stylistic noise, /review enforces strict comment calibration,
empirical verification, and embedded systems invariants.
Operating Modes
Determine the mode from how /review is invoked:
- Mode A — Self-Review Gate (Local Pre-Upload /
/crankSubagent):- Trigger: Bare
/reviewon local uncommitted/unpushed changes, or invoked by/crankbefore./gh pr create/./gh pr ready. - Behavior: Inspects
git diff origin/main...HEAD(or working tree diff), runs targeted verification, and emits a structured verdict (PASSorNEEDS_FIXES) with actionable file/line findings so the author agent fixes every[bug],[embedded/size],[api], and[test]issue before a human reviewer ever sees the CL.
- Trigger: Bare
- Mode B — Gerrit CL Reviewer (
/review <cl>or Reviewer Workflow):- Trigger:
/review <cl_id_or_url>(or bare/reviewwhen reviewing a teammate's CL). - Behavior: Fetches CL context via
./gh pr view <cl> --commentsand./gh pr diff <cl>, evaluates the patch, and stages calibrated inline comments using./gh pr comment <cl> --path <file> --line <line> -m "<msg>" --draft. Never publishes comments live without--draftunless explicitly instructed by the user.
- Trigger:
Comment Calibration & Anti-Slop Rules
Every comment or finding MUST pass these calibration filters:
- Zero Praise / Filler Comments: Never post inline comments that merely say "Nice refactor!", "Looks good", or restate what the code does.
- Verify Before Accusing: Before claiming a symbol is undefined, a header
is missing, or a lock is unlocked, verify with
code_search,view_file, orbazelisk test. Zero hallucinated compiler errors. - Mandatory Severity Prefix: Every inline finding must begin with one of:
[bug]— Correctness, memory safety, race condition, lifetime/dangling reference, or unhandled error status.[embedded/size]— Dynamic heap allocation in a no-alloc module, unbounded recursion/stack frame (>256B buffer on stack), ISR-unsafe blocking call, or unnecessary template/code-size bloat.[api]— Public header C++17 incompatibility, missing#pragma once, symbol outsidepwnamespace, unprefixed macro, or Doxygen mismatch.[build]— MissingBUILD.bazel/BUILD.gn/CMakeLists.txtparity, missing visibility/deps, or broken negative compilation test (PW_NC_TEST).[test]— Missing unit test for new behavior/bugfix, unfocused multi-scenario test, or use of rawEXPECT_EQ(status, pw::OkStatus())instead ofPW_TEST_EXPECT_OK.nit:— Optional, minor readability or naming polish.
- Nit Budget (Max 3): Emit at most 3
nit:comments across the entire CL. If there are blocking[bug]or[embedded/size]issues, prioritize those and drop minor nits entirely. - Respect Formatter Authority: Never leave inline comments about whitespace,
line wrapping, or include sorting that
./pw formathandles automatically. - Changed Lines Only: Review the diff, not the file. Never stage an inline comment on unchanged code; put out-of-diff observations, if any, in the summary (Mode A) or in one change-level draft (Mode B).
- Cite the Rule: When a finding rests on a Pigweed rule, link it as a
pigweed.dev URL, e.g.
https://pigweed.dev/code_reviews.html#docs-code-reviews-checklist-<aspect>orhttps://pigweed.dev/style/<page>.html#<anchor>.
Step-by-Step Review Workflow
Step 1: Gather Context & Diff
- For Mode A (Local Self-Review):
git log -1 --format=%B HEAD git diff origin/main...HEAD - For Mode B (Gerrit CL
<cl>):
If the CL links a Buganizer issue (./gh pr view <cl> --comments ./gh pr diff <cl>Bug: b/<id>orFixed: b/<id>), inspect it via./gh issue view <id>to verify the change actually solves the stated problem. - Hard stops: If no target resolves (no
HEADahead oforigin/mainand no CL given), or the CL isMERGEDorABANDONED(./gh pr view <cl> --json state), stop and say why without dispatching anything.
Step 2: Dispatch Reviewer Subagents & Merge Findings
You coordinate: you do not read changed source yourself when a reviewer subagent handles the file. Work from the diff gathered in Step 1.
Map files to reviewers. A row applies only if a changed file matches it and the agent is enabled (its definition file exists and
git grep -l "^disabled: true" -- <definition file>prints nothing).Files matching Agent ( name:)Definition file *.c,*.cc,*.h,*.inccpp-reviewer.agents/agents/cpp_reviewer.md*.py,*.pyipython-reviewer.agents/agents/python_reviewer.md*.rsrust-reviewer.agents/agents/rust_reviewer.mdpw_ghish/**go-ghish-reviewer.agents/agents/go_ghish_reviewer.mdpw_ghish/**,ghghish-security-reviewer.agents/agents/ghish_security_reviewer.md- Cut one patch per reviewer without reading it.
out=$(mktemp -d), thengit diff origin/main...HEAD -- <its files> > "$out/<agent>.patch"(Mode B: filter the./gh pr diff <cl>output the same way). Dispatch every selected reviewer in one parallel step, each as a read-only background subagent. Give each the checkout path, its patch path, its file list, and this instruction verbatim:
Read and follow
.agents/skills/review/reviewer_workflow.md. Return findings in the Finding format defined there and nothing else.Wait for every reviewer to finish before merging.
- Merge. Render findings with the severity prefixes above as mapped in
reviewer_workflow.md:blockingandshould-fixbecome[<tag>] <issue> (<rule>),nitbecomesnit: <issue>. Deduplicate same file/line/rule; order by severity withscope: in-diffbeforeout-of-diff. Never soften, reword, or drop a reviewer'sblockingfinding. Apply the nit budget to the merged set. - Fall back for the rest. Files matching no row, or only rows whose agent is disabled or absent, get the inline audit below. Report which files each reviewer handled and which fell back.
Fallback: inline audit
For files no enabled reviewer handled, read the full surrounding file context
(view_file) for modified hunks—never review a diff hunk in isolation. Check
these six Pigweed dimensions:
1. Embedded Memory & Execution Constraints
- Zero Dynamic Allocation: Core Pigweed C++ modules disallow
new,delete,std::vector,std::string,std::map, andstd::function. Verify fixed-capacity alternatives (pw::InlineString,pw::Vector,pw::Function,pw::Alloctorinjection) are used. - Stack & Code Size (
pw_bloat): Flag large stack allocations (e.g., byte arrays >256 bytes on MCU stacks), accidental<iostream>/<sstream>includes, or heavy template instantiations on hot paths. - ISR & Thread Safety (
pw_sync): CheckInterruptSpinLockvs.Mutexusage in interrupt-callable contexts, lock annotations (PW_GUARDED_BY,PW_EXCLUSIVE_LOCKS_REQUIRED), andBorrowablelifetimes.
2. C++17 / Rust (pw_kernel) / Python Idioms
- C++17 Compatibility: Public headers (
public/.../*.h) must compile cleanly with-std=c++17. C++20 enhancements must be guarded with__cplusplusorPW_CXX_STANDARD_IS_SUPPORTED(20). - Error Handling: Recoverable errors must use
pw::Status/pw::Result<T>(no C++ exceptions). Fatal invariant violations must usePW_CHECK/PW_ASSERT. pw_async2(if applicable): VerifyPend(Context&)implementations properly register aWaker(PW_ASYNC_STORE_WAKER) whenever returningPending(), and never hold locks across yield points.- Rust /
pw_kernel(if applicable): Verifyno_stdcompliance, minimalunsafeblocks with explicit// SAFETY:justifications, and panic-free error propagation. pw_ghish(if applicable): Forpw_ghishor./ghchanges, launch thepw_ghishsubagent reviewers (.agents/agents/go_ghish_reviewer.mdand.agents/agents/ghish_security_reviewer.md).
3. Build System Parity (Bazel + GN + CMake)
- If a source file (
.cc), public header (.h), or dependency was added or removed inBUILD.bazel, verify whether the module also maintainsBUILD.gn(andCMakeLists.txt). Missing GN parity is one of the top causes of Pigweed CQ breakages!
4. Test Rigor & Negative Compilation Tests
- Every bugfix or behavior change requires a focused
pw_unit_test(*_test.cc) test case; applydocs/sphinx/code_reviews.rst#docs-code-reviews-checklist-tests. - Assertions on
pw::Status/pw::Resultshould usePW_TEST_EXPECT_OKandPW_TEST_ASSERT_OK. - Template/constexpr API constraints should include
PW_NC_TESTnegative compilation tests where appropriate.
5. Documentation & Commit Hygiene
- Public API additions/changes in
public/headers must update Doxygen comments and module.rstdocs (docsskill conventions). - Commit message must follow
<module>: <Imperative subject <=72 chars>, explain the why, includeBug: b/<id>orFixed: b/<id>, and preserveChange-Id:.
6. AI Artifacts
- If the change touches
.agents/,AGENTS.md,GEMINI.md, orCLAUDE.md: readdocs/sphinx/style/ai_artifacts.rstin full at review time and apply its review checklist; cite findings ashttps://pigweed.dev/style/ai_artifacts.html#<anchor>(for example#docs-pw-style-ai-artifacts-review-checklist).
Step 3: Empirical Check (When Local Workspace Matches)
When running in Mode A (or Mode B from a worktree allocated with
./gh wt use review-<cl> --cl <cl> --json; never ./gh pr checkout in a
checkout you did not allocate):
- Run hermetic tests for modified modules:
bazelisk test --noshow_progress --noshow_loading_progress //<module>/... - Check formatting:
./pw format --check
Step 4: Output Verdict & Stage Drafts
In Mode A (Self-Review Gate):
Return a structured markdown report:
- Verdict:
PASS(zero[bug],[embedded/size],[api],[build], or[test]findings) orNEEDS_FIXES. - Blocking Findings: File, line, severity tag, and exact fix required.
- Nits (<= 3): Optional minor improvements.
- Coverage: which files each reviewer subagent handled, which fell back
to the inline audit, and any
scope: out-of-diffobservations.
In Mode B (Gerrit CL Reviewer):
- Check existing comments (
./gh pr view <cl> --comments) so you do not duplicate feedback already left by another reviewer. - Stage each
in-difffinding as a private Gerrit draft:
Put./gh pr comment <cl> --path <file> --line <line> -m "[<tag>] <concise explanation and concrete suggestion>" --draftout-of-difffindings, if any, in one change-level draft (./gh pr comment <cl> -m "<summary>" --draft). Never vote. - Present a summary table of all staged
--draftcomments to the user, along with the command to publish them once reviewed (./gh pr review <cl> --publish).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

