agentleFS
tenforward-audiotenforward-audio/swe-harness/.agents/skills/review-project-change/SKILL.md

review-project-change

Skill0 starsChanged 3 months agoMITView on GitHub

Technically review a defined candidate commit, branch, worktree, or patch without changing product behavior, tracking state, branches, or external systems. Use for pre-integration code review, candidate-versus-base comparison, verification evidence, and actionable findings. Do not use for general investigation, implementation, lifecycle transitions, acceptance, or integration.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 6 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    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
  1. Distinctive:

    Only the active agent responsible for the user-facing outcome applies this mode.

    line 12
  2. Distinctive:

    Read the root AGENTS.md, relevant policies, owning implementation, and existing tests.

    line 19
  3. Distinctive:

    If the candidate cannot be pinned or its base is ambiguous, report that uncertainty instead of reviewing an inferred target.

    line 25
  4. Distinctive:

    Keep the review read-only.

    line 28
  5. Distinctive:

    Run only non-mutating focused checks that materially improve confidence.

    line 42
  6. Distinctive:

    Do not infer compatibility obligations from repository history alone.

    line 45

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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.

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

Review a Project Change

Execution mode

delegate-readonly

Only the active agent responsible for the user-facing outcome applies this mode. A worker already delegated a review executes its assigned scope inline and must not dispatch another subagent. If the scope is too broad or blocked, return the evidence and a proposed split to the active agent.

Pin the review target

  1. Read the root AGENTS.md, relevant policies, owning implementation, and existing tests.
  2. Resolve the exact candidate commit or supplied patch and its intended common base. A moving branch or worktree name is not a stable review target.
  3. Inspect candidate commits, changed paths, the complete diff from the common base, and interaction with current canonical state.
  4. If the candidate cannot be pinned or its base is ambiguous, report that uncertainty instead of reviewing an inferred target.

Keep the review read-only. Do not edit files, queues, boards, branches, commits, worktrees, dependencies, releases, or external systems.

Review risk and evidence

  • Prioritise correctness, security, data loss, public compatibility, concurrency, operational failure behavior, preservation guarantees, and missing tests.
  • Trace changed inputs, state transitions, failures, and outputs through the owning boundary. Review important success, invalid-input, failure, preservation, and supported-compatibility paths.
  • Audit changes to root instructions and .agents/ separately. Identify stale queue, board, policy, or evidence copies that must not displace the coordinating checkout's canonical state.
  • Run only non-mutating focused checks that materially improve confidence. Distinguish automated results, manually observed behavior, unobserved manual acceptance, and checks that could not be run.
  • Do not infer compatibility obligations from repository history alone. Tie any obligation to a supported release, current consumer, authoritative data set, staged rollout, or explicit constraint.

Return a review decision aid

  • Lead with actionable findings ordered by severity. For each finding, name the concrete failure scenario, affected boundary, evidence, and safest fix.
  • Say plainly when no material finding is present and record residual risks or unobserved evidence.
  • Identify the exact reviewed commit, common base, checks run, and whether the candidate changed after review began.
  • Return the evidence without accepting, integrating, or changing lifecycle state. Use manage-project-work for an explicit acceptance or return decision, and integrate-reviewed-change for a separately authorised integration.

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.