Content
77%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 analysis workflow with exact field names, decision rules, and a copy-paste output template. Its weaknesses are token efficiency — the same evidence-handling rules are restated several times and a paragraph about the hosted workflow's infrastructure does not instruct the agent — and bundle wiring, where the primary content reference points outside the skill and half the bundled scripts are never mentioned.
Suggestions
Consolidate the repeated 'missing results are not passes' rule into a single statement in the evidence section and reference it from the other sections instead of restating it four times.
Remove or move the hosted-workflow failure-notice/deduplication paragraph (lines 16-23) — it describes other jobs' behavior, not instructions for the agent executing this skill.
Fix the MAUI CI facts reference: either move the document inside the skill bundle (e.g. references/maui-ci-facts.md) or remove the link, and mention the other bundled scripts (Merge-TestVisualsIntoComment.ps1, Publish-TestVisualAssets.ps1) where relevant so they are discoverable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and mostly prescriptive domain rules, but repeats the same rule several times — "Never treat missing current results as passing" (line 273), "do not... treat absent, skipped, or filtered tests as passes" (line 118), "Results missing from the bundle are not proof that CI has no results" (line 44), "Zero failed tests is not zero test results" (line 52) — and inlines an infrastructure paragraph about the hosted workflow's failure-notice job (lines 16-23) that describes other jobs' behavior rather than instructing the agent. Not 4: the repetition and the non-actionable infrastructure paragraph are more than minor trimmable instances; not 2: the bulk is justified, non-redundant domain guidance rather than padding or explanations of concepts Claude already knows. | 3 / 5 |
Actionability | Quotes: exact context fields to read ("pr.headRefOid, pr.baseRefName, scope.diff, builds, and failures", "history.pipelines", "buildInsights.reports"), a four-row attribution table with required evidence, an exact budget ("at most 250 words"), exact badge parameters ("style=flat-square, labelColor=30363d, and blue 1f6feb"), and a complete copy-paste markdown template with literal HTML entities. As an instruction-only skill the guidance is fully executable, covering the common cases (no failures / pending / no results / missing context). | 5 / 5 |
Workflow Clarity | The five sections form an explicit gated sequence — check for results first (with stop conditions per failure mode, e.g. "If context cannot be found or read, stop with a short report"), read evidence with verification checkpoints ("Verify current builds belong to the captured PR head", "Inspect collection errors and report truncated flags"), compare history, classify via decision table, then publish exactly once. Error-recovery branches and validation checkpoints are present throughout. Not 4: checkpoints are explicit and systematic, not merely 'most' present. | 5 / 5 |
Progressive Disclosure | The body has clear section structure, but the only content reference — "consult the pipeline, data-source, and device-test sections of [MAUI CI facts](../../docs/maui-ci-facts.md)" — points two directories outside the skill bundle and does not resolve, and the bundle's other scripts (Merge-TestVisualsIntoComment.ps1, Publish-TestVisualAssets.ps1 and their tests) are never referenced from the body, leaving them undiscoverable. All detailed evidence-handling rules are inlined in one ~280-line file. Not 4: the broken external path and orphaned bundle files are more than minor organization gaps; not 2: sections are well-organized and the one in-body script reference (Gather-TestFailureContext.ps1) is clearly signaled. | 3 / 5 |
Total | 16 / 20 Passed |