Content
86%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 efficient, tightly organized instruction skill: an exact output template, terse decision rules with explicit escalation criteria, and a well-scoped untrusted-input policy. The main improvement would be naming the concrete commands used to fetch check/review/mergeability data and clarifying how the watched-PR list is discovered.
Suggestions
Add the concrete data-fetch commands, e.g. `gh pr view --json checks,reviews,mergeable,state` or `gh pr checks <number>`, so the triage step is copy-paste executable.
Clarify where the list of watched PRs comes from (loop config or state file) and in what order to process them.
Optionally show one tiny filled-in example of the per-PR report so edge cases like 'absent/unknown' checks are unambiguous.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence: no explanations of what PRs, CI, or mergeability are; the template and rules are terse and every line is a concrete decision criterion, so every token earns its place. | 5 / 5 |
Actionability | The exact output template and concrete decision rules (e.g. 'Zero checks... means absent/unknown', 'idle >4 days -> human handoff') are highly actionable, but no specific data-fetch commands (e.g. gh pr checks, gh pr view --json) are given, leaving a minor execution gap. | 4 / 5 |
Workflow Clarity | The per-PR action is unambiguous with a full report template and explicit escalation checkpoints, but how watched PRs are enumerated and where their state lives is unstated, leaving minor gaps in the sequence. | 4 / 5 |
Progressive Disclosure | Under 50 lines with no need for external references, organized into clear sections (Per-PR Output, Rules, Untrusted input); the single external link is one level deep and the generated untrusted-input block is well delimited. | 5 / 5 |
Total | 18 / 20 Passed |