Content
75%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.
The body is an executable, well-sequenced PR-lifecycle workflow with concrete GitHub and GitLab commands and a clear re-poll feedback loop. It is concise and actionable, with only minor gaps around commit/push validation and PR-number capture.
Suggestions
Add an explicit validation checkpoint after commit/push (e.g. check exit status before opening the PR) to satisfy the destructive-operation feedback-loop guidance.
Turn the PR_NUMBER capture from `gh pr create` / `glab mr create` into actual code (e.g. parse the URL) rather than a comment hint.
Consider extracting the per-platform polling block into a bundled script under scripts/ to slim the inline SKILL.md body.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and purposeful — concrete bash commands with minimal prose and no padding about what a PR or git is; the duplicated per-platform polling logic and the 'Important Notes' recap are minor trims that could be made. | 4 / 5 |
Actionability | Largely executable: complete bash for platform detection, branch, commit, push, PR/MR creation, polling, fetching and replying to comments across both GitHub and GitLab; minor gaps like capturing PR_NUMBER 'from the URL last path segment' are left as comments rather than code. | 4 / 5 |
Workflow Clarity | A clearly numbered 7-step sequence (Branch → Commit → Push/PR → Poll → Present → Act → Re-poll) with a feedback loop back to Step 4 and an explicit timeout; git operations lack explicit success validation before proceeding, so it stops short of 5. | 4 / 5 |
Progressive Disclosure | No bundle files exist and the skill is a self-contained workflow with clean section headers and numbered steps; the long inlined per-platform polling blocks could plausibly live in script files, which is the minor organization gap keeping it from 5. | 4 / 5 |
Total | 16 / 20 Passed |