immutable-pr-review
Skill4 starsChanged 2 months agoApache-2.0View on GitHub
Review a requested PR, patch, or working-tree diff read-only for concrete failures and contract gaps. Use surrounding source and tests; not a demand for a revision packet or authority to edit.
In this repository81 other files it gives its agents.
- agent-cli-designSkill
- agent-tool-boundariesSkill
- behavior-first-implementationSkill
- bounded-prototypeSkill
- bounded-pr-sliceSkill
- building-fdd-validationSkill
- context-efficiencySkill
- cross-language-contractsSkill
- cross-platform-agent-instructionsSkill
- design-interviewSkill
- diagnosis-loopSkill
- domain-languageSkill
- edge-cloud-syncSkill
- handoff-continuitySkill
- memory-context-hygieneSkill
- module-designSkill
- orchestrateSkill
- owner-decision-gateSkill
- parallel-portfolioSkill
- pr-gate-loopSkill
- pyo3-async-lifecycleSkill
- pyo3-buffer-performanceSkill
- pyo3-free-threadingSkill
- pyo3-ownership-errorsSkill
- pyo3-wheel-releaseSkill
- python-api-typingSkill
- python-async-concurrencySkill
- python-ci-releaseSkill
- python-data-boundariesSkill
- python-packaging-environmentsSkill
- python-performanceSkill
- python-pytestSkill
- python-reviewSkill
- python-test-designSkill
- python-threading-workersSkill
- python-workflowSkill
- react-components-stateSkill
- react-performanceSkill
- repository-groundingSkill
- rust-api-designSkill
- rust-async-concurrencySkill
- rust-build-diagnosticsSkill
- rust-ci-releaseSkill
- rust-memory-layoutSkill
- rust-nextestSkill
- rust-numerics-simdSkill
- rust-performanceSkill
- rust-protocol-codecsSkill
- rust-python-bindingsSkill
- rust-reviewSkill
- rust-storage-durabilitySkill
- rust-test-designSkill
- rust-unsafe-ffiSkill
- rust-workflowSkill
- rust-workspace-dependenciesSkill
- shadcn-react-componentsSkill
- shadcn-svelte-componentsSkill
- simplify-existing-codeSkill
- source-verificationSkill
- svelte-components-runesSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 21
Look for reachable incorrect behavior: ownership/lifetime errors, cancellation and cleanup gaps, lost or duplicated data, invalid …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Evidence-Based PR Review
The directory name remains for existing callers. Review the actual requested change without requiring a special packet, a clean unrelated worktree, a particular agent role, or a revision ledger.
Establish the scope
Identify the PR, supplied patch, or local changes under review. Read applicable instructions, the relevant diff, surrounding implementations, callers, and tests. Inspect missing context with available read tools. An optional code index can help trace dependencies, but source remains the basis for consequential findings.
For a working-tree review, include the requested uncommitted changes and preserve unrelated work. If unrelated edits make attribution ambiguous, state that limitation rather than resetting them. For a patch-only review without surrounding code, narrow the conclusions accordingly.
Remain read-only by default. Focused execution is appropriate when authorized, safe, and unlikely to alter user data; otherwise reason from source and explain what remains untested. Do not edit, post, approve, commit, push, or merge without separate scope.
Follow the failure path
Look for reachable incorrect behavior: ownership/lifetime errors, cancellation and cleanup gaps, lost or duplicated data, invalid contracts, security boundaries, silent error suppression, and missing consumer coverage. Use the relevant language specialist when needed. Assess performance claims against measurements rather than intuition alone.
A useful finding gives its location, triggering input or state, why the behavior fails, its impact, and the smallest correction or regression test. Direct source reasoning can establish a defect even when reproducing it is unsafe or unavailable; label that evidence honestly. Do not demand a runtime exploit or an observed production failure.
Separate introduced defects from pre-existing issues. Report a root cause once. Keep preferences, speculative hardening, and nits distinct from blockers. A missing test is most meaningful when tied to a specific unprotected behavior.
Return a bounded result
Lead with actionable findings, then confidence, validation performed, and material limits. A clean review means no supported issue was found in the inspected scope—not certification, passing CI, or merge permission. If the change moves materially during review, state what was inspected and reassess affected work rather than claiming coverage of unseen code.
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.

