Content
100%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
An exemplary instruction-only skill body: it specifies exact gh/git commands, a clear classification scheme for CI failures and review comments, explicit validation checkpoints within a feedback loop, and safety guardrails — all in a lean ~27 lines with no padding. Nothing material is missing at this scope.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every line is a directive, decision rule, or guardrail ("Don't poll with sleeps", "A second failure of the same shape is not a flake; treat it as real", "Never plain force-push a shared branch") with zero re-teaching of concepts Claude already knows, matching the lean anchor 5 rather than anchor 4's "minor instances of over-explanation". | 5 / 5 |
Actionability | The body is fully executable with exact commands and flags throughout — `gh pr view <number> --json number,headRefName,baseRefName,mergeable,reviewDecision,statusCheckRollup`, `gh pr checks <number> --watch --fail-fast`, `gh run view <run-id> --log-failed`, `gh api repos/{owner}/{repo}/pulls/<number>/comments`, and `push with --force-with-lease` — covering the common cases for each loop step. | 5 / 5 |
Workflow Clarity | A numbered six-step loop with explicit validation feedback — re-watch CI after every push, "rerun the local checks" after a rebase, a rerun-once flake policy with escalation to real failure — plus a defined exit condition ("until the exit condition holds or two consecutive rounds make no progress"), matching the anchor-5 feedback-loop pattern. | 5 / 5 |
Progressive Disclosure | The body is under 50 lines with no need for external references (none exist in the bundle) and is cleanly organized into focused sections (intro contract, "## Loop", "## Guardrails"), which per the rubric's simple-skill guideline merits a 5. | 5 / 5 |
Total | 20 / 20 Passed |