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 skill body is a highly actionable, clearly sequenced workflow with per-step gates, concrete scripts, and well-structured one-level-deep references. Its weaknesses are verbosity in several qualifying sentences and one broken local reference (sub-agents/frame-worker.md).
Suggestions
Resolve or remove the missing `sub-agents/frame-worker.md` reference — the file is cited in Step 5 and the Quick Reference table but no `sub-agents/` directory exists, which breaks a signaled one-level-deep reference.
Tighten the densest paragraphs (Step 0 opening rule, Step 3 review-loop, Step 4 catalog-search) by splitting stacked parenthetical exceptions into short separate lines or a small list, cutting tokens without losing the gates.
Consider lifting the per-step **Gate:** lines into a compact checklist near the top so the gate sequence is scannable without reading each step.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and assumes Claude's competence on the domain, but it is markedly verbose — many sentences are heavily qualified with parenthetical exceptions and inline rationale (e.g. the Step 3 review-loop paragraph and the Step 0 opening-rule sentence run long with stacked clarifications), so it sits between efficient and mostly-efficient with notable tightening possible. | 3 / 5 |
Actionability | Every step supplies concrete, executable commands — exact `node <SKILL_DIR>/scripts/*.mjs ...` invocations with flags, `npx hyperframes` calls, and `gh api` snippets covering the common cases — so guidance is copy-paste ready and specific. | 5 / 5 |
Workflow Clarity | The process is a clearly numbered Step 0→6 sequence with an explicit **Gate** at the end of every step, user-gated steps called out, error/feedback handling (re-dispatch that one frame only, one retry max; surface stderr and stop; fix cheapest safe edit then rerun), and the known-false-positive guidance for `check`, hitting the explicit-validation + feedback-loop anchor. | 5 / 5 |
Progressive Disclosure | Structure is strong: an overview SKILL.md with a 'Quick Reference' Read-table mapping each reference to its step, local references and scripts one level deep, and design/motion rules correctly deferred to sibling-skill files; it is not a 5 because one referenced local file — `sub-agents/frame-worker.md` (cited in Step 5 and the Read table) — does not exist, so a signaled reference is broken. | 4 / 5 |
Total | 17 / 20 Passed |