Content
50%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill provides a reasonable framework for duplicate issue detection with good examples and clear comment formatting rules. However, it leans toward describing what to do rather than providing fully executable, systematic guidance—the search strategy is underspecified and the workflow lacks validation checkpoints. The examples are helpful but verbose, and the promised GitHub Actions automation is never delivered.
Suggestions
Make the search strategy more concrete by providing specific gh search query patterns (e.g., exact error message quoting, label filters, date ranges) rather than abstract descriptions like 'search for similar symptoms'.
Add a validation step after posting the comment (e.g., `gh issue view <number> --comments` to verify the comment was posted) and an explicit exit/do-nothing path in the workflow for low-confidence cases.
Move the detailed examples to a separate EXAMPLES.md file and reference it from the main skill, or condense them significantly—Example 2 and 3 could each be half their current length.
Either add the GitHub Actions automation content mentioned in the intro or remove that claim to avoid misleading expectations.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is reasonably efficient but includes some content Claude could infer (e.g., the detailed comparison criteria like 'same area of code?' and the low confidence indicators are somewhat obvious). The examples section, while useful, is quite lengthy and could be more compact. | 2 / 3 |
Actionability | The skill provides concrete gh CLI commands and clear comment formatting rules, but the core search strategy is vague—it says 'use multiple strategies' and 'search for similar error messages' without providing executable search patterns or a systematic approach. The gathering step is descriptive rather than instructive. | 2 / 3 |
Workflow Clarity | The four-step workflow is clearly sequenced and the confidence threshold (90%+) is a good checkpoint. However, there's no validation or feedback loop—no step to verify the comment was posted correctly, no guidance on what to do if search returns too many or too few results, and no explicit 'stop and do nothing' exit path integrated into the workflow steps. | 2 / 3 |
Progressive Disclosure | The content is well-structured with clear headers and sections, but the three detailed examples make the file quite long. The examples could be in a separate file, and the GitHub Actions automation mentioned in the intro is never addressed, which is a missed opportunity for a reference link. | 2 / 3 |
Total | 8 / 12 Passed |