Content
63%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 a strong, executable playbook: six conflict patterns each pair concrete detection commands with specific resolution strategies, an explicit audit sequence, and post-resolution verification. Its main weaknesses are redundancy (repeated reset commands, a very long output template) and a monolithic single-file layout that inlines reference material.
Suggestions
Consolidate the three repeated SPM reset instruction blocks (Pattern 4, the output template, and the commands reference) into one canonical reset procedure.
Move the full output-format template and the SPM/Xcode command reference into references/ files (e.g., references/output-format.md, references/commands.md) and keep only a skeleton plus pointers in SKILL.md.
Trim the 'Invocation Examples' section (it duplicates the description's triggers) and remove the padded `products: [...], dependencies: [...]` boilerplate from the Swift examples.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient, command-dense content, but the destructive reset (rm -rf .build, rm Package.resolved, swift package resolve) appears three times (Pattern 4, the output template's 'Resolution Commands', and 'Common SPM Commands Reference'), and the ~130-line worked output template plus 'Invocation Examples' restate trigger guidance already in the description. Not 2 because most sections are concrete and unpadded; not 4 because the duplication and placeholder Swift boilerplate (products: [...]) are noticeable. | 3 / 5 |
Actionability | Highly executable guidance: exact commands (swift package show-dependencies --format json, xcproject inspect --root ., a jq duplicate-pin pipeline with a no-jq fallback) and specific resolution edits per pattern. Minor gaps: Swift manifest snippets use [...] placeholders, and swift package diagnose-api-breaking-changes is offered as a generic 'check for issues/security advisories' tool, which slightly mischaracterizes it. Below anchor 5 for these gaps; well above 3. | 4 / 5 |
Workflow Clarity | The Audit Process gives a clear Step 1→3 sequence (gather info, identify conflicts, analyze graph) and a Verification section closes the loop (clean build, swift test, grep warnings), so the destructive resets are followed by validation — the cap-3 rule for destructive operations without validation does not apply. Not 5 because the detect→fix→re-detect feedback loop is implicit and the 'nuclear option' cache wipe carries no pre-check caution. | 4 / 5 |
Progressive Disclosure | A single ~480-line file with good section headers but no bundle: the full output-format template and the command reference are inline content that would fit better in separate reference files, and the only pointer is the external <plugin-root>/tools/xcproject. Matches anchor 3 (structure present, content that should be separate is inline); not 2 because sections are well-headed and navigable; not 4 because at this length the all-inline placement exceeds a minor gap. | 3 / 5 |
Total | 14 / 20 Passed |