Content
92%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.
A well-engineered loop skill: executable commands, an explicit sequenced workflow with validation and feedback loops, and clean one-level-deep progressive disclosure via real bundle files. The only slack is minor philosophical prose in the principle section.
Suggestions
Tighten the 'Principle: you are pragmatic, not obedient' section — drop rhetorical lines like 'You are having a technical conversation, not winning an argument' that restate the already-clear bar.
The Fix/Push back/Verify bullets are useful but could be compressed into a tighter decision table to reduce token weight in the hot path of the loop.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Largely lean and assumes Claude's competence (no preamble explaining gh, PRs, or bots), but the 'pragmatic, not obedient' principle section carries some philosophical prose ('You are having a technical conversation, not winning an argument') that could be trimmed without losing guidance. | 4 / 5 |
Actionability | Fully executable: exact gh REST endpoints, copy-paste bash blocks, helper scripts invoked with concrete arguments, and ready-made accepted/rejected reply templates covering the common cases. | 5 / 5 |
Workflow Clarity | A clearly sequenced 6-step loop with explicit validation checkpoints — empty-array check routes to termination, 'verify before trusting' against real code, and step 6 polling with timeout/re-run/reported-stall feedback loops for error recovery on this batch operation. | 5 / 5 |
Progressive Disclosure | SKILL.md is a focused overview with well-signaled one-level-deep references — helper scripts in scripts/ and optional gotchas/resolution mechanics in references/extras.md, all real files enumerated in a Files section and linked inline; nothing nested deeper. | 5 / 5 |
Total | 19 / 20 Passed |