Content
65%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A highly actionable, well-sequenced moderation workflow with real gh commands, but it is held back by heavy verbatim duplication of the notice block, missing post-edit verification for destructive/batch operations, and a monolithic structure with no progressive disclosure.
Suggestions
Factor the code-of-conduct notice text into a single referenced snippet (e.g., a notice template in a references/ file) instead of repeating it verbatim six times across the title, body, and comment sections.
Add explicit validation/verification after destructive edits — e.g., re-run 'gh issue view' to confirm the title/body changed and report confirmation — so the workflow clears the destructive/batch operation bar for workflow_clarity.
Split the per-violation procedures (title, body, comments) into a reference file and keep SKILL.md as an overview, or at minimum de-duplicate the Notice/Principles/Notes sections that restate the same title-replacement and verbatim-notice guidance.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | It avoids explaining concepts Claude already knows, but the full code-of-conduct notice is duplicated verbatim six times (three markdown blocks plus three bash implementations) and the Notes/Important Principles sections restate earlier guidance, so it is mostly efficient but could be tightened rather than fully lean. | 2 / 3 |
Actionability | Provides fully executable commands with concrete arguments — 'gh issue view $1 --repo $0 --comments --json comments', 'gh issue edit $1 --repo $0 --title "…"', and 'gh api --method PATCH /repos/$0/issues/comments/{comment-id}' — plus a concrete title-replacement example, making it copy-paste ready. | 3 / 3 |
Workflow Clarity | Steps are well sequenced with a title-meaningfulness checkpoint, but destructive in-place edits (title, body, comments) and batch comment sanitization lack any post-edit success verification or re-fetch, so the rubric's destructive/batch cap holds this at 2 rather than 3. | 2 / 3 |
Progressive Disclosure | No bundle files exist and the skill is a single >50-line monolithic file with no external references, yet it is organized into clear sections; the large repeated notice blocks are content that could be factored out, fitting the 'some structure but content that should be separate is inline' anchor rather than the poor-organization anchor. | 2 / 3 |
Total | 9 / 12 Passed |