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.
The body is a strongly actionable, well-gated multi-mode workflow with excellent validation checkpoints and a genuinely load-matched reference structure. Its main weakness is redundancy — several rules are restated two or three times and a large output template is inlined, which inflates token cost without adding information.
Suggestions
State the "new issue discovered mid-fix → stop, surface, add/defer/comment" rule once (e.g., in Critical rules) and reference it from the Linear protocol and Phase 3 instead of repeating it three times.
Move the findings-doc template (~75 lines) into references/findings-doc-template.md and load it only in the no-task case, consistent with the skill's own context-hygiene table.
Drop the entry-point Behavior column prose that duplicates the mode-workflow sections, or fold the entry-point table into the workflows to remove the overlap.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and operational with no padding on concepts Claude already knows, but the "new issue mid-fix → stop/surface/ask (a)(b)(c)" rule is stated three times (Linear protocol, Phase 3 step 12, Critical rule 1b), PR-convention deferral is stated twice (Phase 6 and Critical rule 7), the entry-point table overlaps the mode-workflow prose, and the ~75-line findings-doc template is inlined — noticeably more than minor trimming. | 3 / 5 |
Actionability | Fully executable guidance throughout: exact MCP calls with arguments (mcp__linear-server__get_issue id="NXC-XXXX"), gh pr view/diff commands with --repo flags, npx nx test ... --testPathPattern invocations, file paths, branch naming conventions, and a parameterized spec helper — copy-paste ready for the common cases. | 5 / 5 |
Workflow Clarity | Six clearly sequenced phases with explicit validation checkpoints (user OK gate before Phase 3 edits, Phase 4–5 test runs and format check), error-recovery feedback (MCP-unavailable fallback to discovery, halt-and-ask on bad task ID, the assert-before-parameterized-spec footgun), and the A–F verification checklist. | 5 / 5 |
Progressive Disclosure | The "Which references to load" table maps each mode to required/optional files, all four referenced bundle files exist and are one level deep, and the body points to specific sections by name. Minor gap: the ~75-line findings-doc template is inlined in SKILL.md where a reference file would better match the skill's own context-hygiene principle. | 4 / 5 |
Total | 17 / 20 Passed |