apply-review-findings
Skill2 starsChanged 15 days agoMITView on GitHub
Use when addressing review comments, bot reports or adversarial findings on a branch or worktree, especially when the target may have advanced since the reviewed revision.
In this repository29 other files it gives its agents.
- AGENTS.mdAGENTS.md
- author-workflow-fanoutSkill
- chain-launcherSkill
- delegateSkill
- design-consultationSkill
- design-fail-closed-gateSkill
- document-releaseSkill
- gistSkill
- goal-promptSkill
- grokbotSkill
- incident-as-codeSkill
- investigateSkill
- learn-captureSkill
- office-hoursSkill
- orion-deep-researchSkill
- oss-loopSkill
- pre-compactSkill
- pre-prSkill
- prove-control-bindsSkill
- prove-deploy-is-liveSkill
- readonlySkill
- reassert-guards-on-rewriteSkill
- reprobe-stale-premiseSkill
- scope-guardSkill
- shipSkill
- tdd-loopSkill
- tools-routerSkill
- triage-fanout-verdictsSkill
- xSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
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 13
Read the repository's contract, checks and applicable prior solutions.
- Distinctive: line 40
If a high-severity declined finding is raised independently again, settle it with a focused reproduction and the controlling contract.
- Distinctive: line 43
If the first fix fails, compare bounded alternatives against the same defect and normal-case controls, retaining rejected alternatives and …
- Distinctive: line 48
Keep a ledger with finding ID/author, reviewed SHA, current SHA, current location, disposition, reason, commands/results, final …
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
Apply review findings
A review is evidence about a snapshot. Re-establish the defect at the current target before changing source or calling a finding resolved.
Before editing
Read the repository's contract, checks and applicable prior solutions. Record the review source, finding IDs, authors, reviewed commit SHA and target. If the reviewed SHA or attribution is unavailable, record unknown and obtain evidence; do not invent it or claim resolution.
Refresh the authoritative target and record the current SHA and PR/branch state. In an offline review, declare the pinned local target and the unverified remote state. A merged PR requires an authorized follow-up vehicle from the current target; a surviving branch is insufficient. Verify write ownership and isolate your work using the repository's normal mechanism. A competing or unknown owner blocks overlapping writes. No particular coordination service is required.
Per-finding decision
Re-read the cited implementation and trace moved code at the current SHA. Choose one disposition with evidence and an attributable reason:
| Disposition | Use when | Evidence or next step |
fix | Defect remains and repair is authorized | Reproduction, minimal repair, final checks and re-read |
already-satisfied | Current behavior already meets the requirement | Trace the fix or move and verify the original failure case |
decline-non-defect | Claim is contradicted or inapplicable to this target | Current code/contract evidence; explain disagreement |
operator/design-gated | Decision exceeds authority or needs a design choice | Preserve behavior and name the required decision |
needs-new-vehicle | Work targets a merged/retired branch or different surface | Identify the current target; no writes until ownership/authority is established |
Staleness is a reason to revalidate every finding, never a reason to discard one. Preserve stronger current guards even if a literal suggestion would weaken them. If a high-severity declined finding is raised independently again, settle it with a focused reproduction and the controlling contract; revise the disposition when the evidence changes. Prior solutions inform the probe, not its verdict. If the first fix fails, compare bounded alternatives against the same defect and normal-case controls, retaining rejected alternatives and their evidence.
Verify and report
Keep a ledger with finding ID/author, reviewed SHA, current SHA, current location, disposition, reason, commands/results, final verification SHA and remaining action. Retain contradictory evidence. Missing or failed checks leave the item unverified. For each fix, re-read the actual defect site after the change and run the repository's relevant checks. Bind results to the tested bytes: before claiming resolution at a commit SHA, run the checks on source materialized from that immutable revision, including all relevant source inputs, without working-tree overlays. A clean Git status alone is insufficient: index flags can hide changed bytes. The worked ledger demonstrates reading the module from a Git object. For an uncommitted repair, retain its patch/tree identity, label the result local/uncommitted and do not claim HEAD contains the fix. This grants no commit authority. Recheck target identity before reporting; if it advances, revalidate affected evidence and record the new SHA. A commit or a stale-thread marker alone is insufficient.
The worked ledger covers current-valid, stale, already-fixed, disputed, inapplicable and policy-gated findings. It also shows a merged-branch change of vehicle. Evidence-backed dispositions may be prepared locally. Posting replies or resolving remote review threads requires separate explicit authorization; never silently mark unverified items resolved.
Prerequisites and authority
Require access to reviewed/current source, review provenance, repository checks and ownership evidence. Git is needed for commit identities; hosted PR state requires authorized access to the host. Missing access remains a stated gap. There are no required sibling skills, providers, private services or lock helpers. This procedure grants no source-write, commit, push, merge, service, credential, installation or external-communication authority.
Retain provenance and the notice with a standalone copy.
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.

