agentleFS
crissmoldovancrissmoldovan/agent-skills/skills/request-blocks-review/SKILL.md

request-blocks-review

Skill2 starsChanged 48 days agoMITView on GitHub

Run the Blocks review, fix and re-review loop on a finished GitHub pull request until its current head is clean. Symptoms: request Blocks review, run Blocks review, fix the Blocks findings, re-review until green. Not for exploratory work or before a pull request exists; the primitives it calls are the blocks skill's.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 6 are distinctive, and 2 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.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Use this as the final code-review gate after implementation and local verification are complete on a GitHub pull request.

    line 12
  2. Distinctive:

    Do not start during exploratory work, before a PR exists, or while implementation and known local failures remain unfinished.

    line 24
  3. Distinctive:

    Never install it silently.

    line 35
  4. Distinctive:

    Resolve the PR: Use the named PR or current branch PR.

    line 43
  5. Very specific:

    Await visibly through blocks: Keep a tracked operation active and surface waiting progress.

    line 48
  6. Distinctive:

    Collect every finding: Read top-level summaries, formal reviews, and all paginated inline comments.

    line 50
  7. Distinctive:

    Verify and push: Run focused and repository-required checks, then push the accepted fixes to the PR branch.

    line 57
  8. Very specific:

    Stop only for a clean current-head review, explicit user cancellation, closed PR, failed, or non-convergence that must be escalated to the …

    line 70

≈ “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.

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

Request Blocks review

Use this as the final code-review gate after implementation and local verification are complete on a GitHub pull request. It owns the review policy and loop; the public blocks skill owns Blocks communication, workspace/session resolution, status classification, and visible waiting primitives.

When to Use

  • Work on a GitHub PR is finished and verified, and it needs Blocks review.
  • “Request Blocks review,” “run Blocks review,” “fix the Blocks findings,” or “re-review until green.”
  • A Blocks review landed and its findings need triage and another review cycle.

Do not start during exploratory work, before a PR exists, or while implementation and known local failures remain unfinished.

Prerequisites

  1. Load the public blocks skill. If unavailable, show and request approval for:

    npx skills add crissmoldovan/agent-skills --skill blocks

    Never install it silently.

  2. Confirm the branch is pushed, the PR is open, implementation is complete, and relevant local and CI checks are green.
  3. Resolve Blocks workspace context through blocks only if REST interaction is needed. GitHub review status remains GitHub-authoritative.

Procedure

  1. Resolve the PR. Use the named PR or current branch PR. Do not create or merge a PR implicitly.
  2. Capture a full baseline. Record the head SHA, request timestamp, and stable IDs for top-level comments, reviews, and paginated inline comments.
  3. Request review once. Post @blocks please review; record its URL.
  4. Await visibly through blocks. Keep a tracked operation active and surface waiting progress. Reactions/help/queued/courtesy messages are nonterminal.
  5. Collect every finding. Read top-level summaries, formal reviews, and all paginated inline comments. Preserve explicit severity; otherwise classify: security/data loss/broken contract = blocker; correctness/crash = high; refactor/quality = medium; style = low.
  6. Apply policy. Fix blocker/high findings. Ask before medium/low judgment calls. For intentional disagreements, reply with rationale rather than changing correct code.
  7. Verify and push. Run focused and repository-required checks, then push the accepted fixes to the PR branch.
  8. Re-baseline and request re-review. A review of an older head never accepts the current head.
  9. Repeat until green. Continue review → findings → fixes → verification → re-review until Blocks reports the current head clean and CI remains green.

Acceptance is not the word "clean". It is a clean verdict for the head you are about to merge, with CI green on that same commit — blocks reports this and names its reason when it refuses. Check the sha, never the check name: a green check can belong to the previous head when the new run has not registered yet, and a verdict for a superseded commit reads exactly like one for this commit.

Stop only for a clean current-head review, explicit user cancellation, closed PR, failed, or non-convergence that must be escalated to the user. Never merge implicitly.

failed is Blocks saying it did not run — logged out, or out of quota. Stop and report its reason; do not re-request in a loop, and never treat it as a pass. A logged-out agent is reconnected only by the workspace owner in the Blocks dashboard.

Usage Examples

Implementation is finished and tests are green. Request Blocks review for this
branch's PR, await visibly, fix blocker/high findings, ask me about judgment
calls, and repeat until Blocks reports the current head clean. Do not merge.
Resume the Blocks review loop on PR 42. Compare against the latest request
baseline, include all inline findings, and do not accept a review of an older
head.

Verification

  • blocks was loaded as the interaction/status/wait authority.
  • Implementation was complete and local/CI checks were green before review.
  • Baseline includes head SHA and stable IDs from all three GitHub surfaces.
  • Wait remained visible and bounded.
  • Every accepted finding was fixed and freshly verified.
  • Re-review targeted the new head after each fix cycle.
  • The final Blocks verdict is clean for the current head and CI is green.
  • No implicit merge occurred.

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.