agentleFS
dativo-iodativo-io/talon/.claude/skills/respond-to-review/SKILL.md

respond-to-review

Skill18 starsChanged 25 days agoApache-2.0View on GitHub

Work through a code-review round on a Talon pull request — classify each finding, reproduce it with a failing test before fixing, prove the fix, and refresh the PR body — on the PR branch only, without merging.

6 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

1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 6 edited versions we read keep these.

  • 1

    Distinctive

    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:

    Classify: before editing: confirmed defect, not a defect (document and pin), already fixed on the branch (name the commit), or out of scope …

    line 10

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

Address the review of $ARGUMENTS on its own branch. Do not merge, do not touch other repositories named in the review (for example talon-full-demo), do not modify the issue or PR unless the user asks.

Per finding

  1. Classify before editing: confirmed defect, not a defect (document and pin), already fixed on the branch (name the commit), or out of scope (say why and propose a follow-up).
  2. Reproduce first. Write the smallest failing test that states the reviewer's claim. When the reviewer gives a shape (cap, estimates, request count, shared or separate sessions, a primary that fails transiently), use exactly that shape. Concurrency tests are deterministic: a fixed burst started behind a sync.WaitGroup or channel barrier, a counting upstream, fixed estimates, no sleeps.
  3. Fix the mechanism, not the symptom. Never restore CI by faking an input (PricingKnown=true in a test that validates pricing, a silent fallback price, a widened allowlist). When an external runtime wraps every Talon denial in its own code (OpenShell renders any deny as middleware_denied), find the underlying Talon code in Talon's evidence or logs and fix that.
  4. Prove it. The new test fails with the fix disabled (temporarily, locally) and passes with it. Keep the atomic semantics the review asked to preserve (for example the session-store reservation) and say how.
  5. Run the ladder (/verify-change) for the packages touched, then the tiers the change crosses. Race detector on for anything concurrent.
  6. Document: LIMITATIONS.md and the reference docs when the guarantee changed; CHANGELOG when behaviour changed.

Findings that are not defects

Pin the behaviour with a named test, state the mechanism (for example one settle per request across streaming and retries), and mark the row "not a defect, pinned by …" in the PR body. Do not change code to satisfy a misreading.

Refresh the PR body

Update, never append-only: the summary table (verdict per finding), changes by commit, regression tests, verification matrix, CI history (commit → failing step → root cause → fixing commit), limitations, migration implications. Write it to a file outside the tree and hand the user gh pr edit <n> --body-file <path>; verify afterwards that the body on GitHub changed.

Handoff

A table with one row per finding: classification, test that proves it, commit, residual risk. State exact commands run and their results, what needs the PR's CI (Docker lanes), and that the PR remains a draft until checks are green. The user decides when to mark it ready and merge.

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.