agentleFS
itsofficialadiitsofficialadi/boundiff/.agents/skills/independent-review/SKILL.md

independent-review

Skill0 starsChanged 26 days agoMITView on GitHub

Perform an independent, read-only review of a commit, branch, pull request, or code path and report reproducible findings. Use for review/audit requests; never silently remediate the target.

14 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. 0 are common, 6 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 10 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Core invariant: review is not remediation: Treat the review target as evidence that must not be modified by the reviewer.

    line 8
  2. Distinctive:

    If the target cannot be identified reliably or unexpected state makes findings ambiguous, stop and report the provenance problem.

    line 14
  3. Distinctive:

    Run negative or mutation testing only on disposable copies with an explicitly bounded target.

    line 20
  4. Distinctive:

    No source edits, auto-fixes, dependency updates, lockfile changes, staging, commits, branch switches, rebases, pushes, or silent …

    line 25
  5. Distinctive:

    No opportunistic edits even when the fix appears trivial.

    line 26
  6. Distinctive:

    Do not treat successful tests as proof that all defects or vulnerabilities are absent.

    line 27
  7. Very specific:

    If a defect warrants remediation, report it and request a new, explicitly authorized remediation task.

    line 29

“Common” is measured: kept by at least 6 in 10 of the 10 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Independent review

Core invariant: review is not remediation. Treat the review target as evidence that must not be modified by the reviewer.

Establish the target

  1. Confirm the repository and exact target (branch/commit/PR, base reference, or files). Record relevant HEAD, working-tree and staged state, and whether the review includes a dirty overlay.
  2. Explicitly distinguish committed source, uncommitted changes, generated files, disposable test mutations, third-party evidence, and assumptions.
  3. If the target cannot be identified reliably or unexpected state makes findings ambiguous, stop and report the provenance problem.

Inspect and test

  1. Review requested behavior, changed paths, callers, tests, error handling, trust boundaries, and potential regressions in proportion to risk.
  2. Reproduce suspicious behavior with narrow, deterministic checks. For tests that may write caches, snapshots, build outputs, files, database state, or Git metadata, use a disposable copy or isolated environment. Do not change the reviewed working tree, index, refs, history, or project configuration.
  3. Run negative or mutation testing only on disposable copies with an explicitly bounded target; never run destructive tests against live or production resources.
  4. Classify findings by severity and likelihood. Support each with affected location, trigger, impact, and evidence. Separate confirmed defects from hypotheses.

Prohibited during review

  • No source edits, auto-fixes, dependency updates, lockfile changes, staging, commits, branch switches, rebases, pushes, or silent remediation.
  • No opportunistic edits even when the fix appears trivial.
  • Do not treat successful tests as proof that all defects or vulnerabilities are absent.

If a defect warrants remediation, report it and request a new, explicitly authorized remediation task. A later task may use $surgical-change or $bug-fix-verification; it does not retroactively change this review into a write-capable operation.

Output

  • Target and provenance: exact object/range and overlay state, if known.
  • Findings: severity, location, reproduction/impact, and confidence, ordered by significance.
  • Verification: checks performed, isolation method, and results.
  • Limitations: unchecked areas, missing environment, and hypotheses.
  • Review state: explicit confirmation that the reviewed target was left unchanged, or a disclosure if this could not be established.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.