Content
81%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 exceptionally actionable, well-sequenced procedural skill body: exact data fields, a rigorous attribution rubric, and a complete output template with validation checkpoints and error-recovery loops. Its main weaknesses are noticeable rule repetition across sections and a bundle whose other scripts are never referenced from the body.
Suggestions
Consolidate the no-results vs collection-failure rules into one canonical passage (e.g., within 'Check for results first') and reference it from the reporting sections instead of restating it.
Briefly mention the other bundle scripts (Merge-TestVisualsIntoComment.ps1, Publish-TestVisualAssets.ps1) or their purpose so the scripts/ directory is fully navigable from the body.
Consider moving the verbatim comment template or the detailed five-run comparison caveats into a reference file to reduce SKILL.md token load.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every sentence carries a substantive rule with no concept-explanation padding, but the no-results-vs-collection-failure distinction is restated three-plus times across sections ("Results missing from the bundle are not proof that CI has no results", "do not say No results available or request /azp run for a collection failure", "Never treat missing current results as passing"), and layout rules are repeated — more than the minor trimming of a 4. | 3 / 5 |
Actionability | Fully executable guidance: exact context field names (pr.headRefOid, pr.baseRefName, scope.diff, builds, failures), a four-row attribution table with required evidence, exact Shields badge URLs with labelColor=30363d, emoji HTML entities, and a complete copy-paste comment template covering the common cases including the no-results shortcut. | 5 / 5 |
Workflow Clarity | A clearly sequenced workflow (check results first → read evidence → compare latest five target-branch runs → classify → produce comment) with explicit validation steps ("Verify current builds belong to the captured PR head", "require complete results for expected Helix work items") and feedback loops for error recovery (stop and ask to restore access and rerun /review tests). | 5 / 5 |
Progressive Disclosure | Good structure with well-signaled one-level-deep references: scripts/Gather-TestFailureContext.ps1 resolves to a real bundle file with its role stated, and the MAUI CI facts reference is section-signaled. Minor gaps: the body never references the other three bundle scripts (Merge-TestVisualsIntoComment, Publish-TestVisualAssets), and the long inline output template plus comparison rules could partially live in references. | 4 / 5 |
Total | 17 / 20 Passed |