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.
A highly actionable, well-sequenced authoring guide with strong validation gates and real, well-signaled bundle files. The main weaknesses are token efficiency — an off-topic question-format section pushes the body against its own budget — and under-use of references/ for the large inline tables and the priming essay.
Suggestions
Trim the "Question Format (FR-UNIVERSAL.QA-FORMAT)" section to only what applies when the skill asks the user about IDE, scope, and behavior during Phase 1, or move it to a reference file — most of its rules govern general question-asking rather than skill creation.
Move the Control Primitives Map table and/or the Priming section into references/ so the SKILL.md body sits comfortably below its own stated 20000-char cap.
Condense the dense "Two validators, two roles" paragraph into a short two-row table (validator vs. limits enforced) to cut tokens and improve scannability.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient (tables, checklists, terse examples), but the "Question Format (FR-UNIVERSAL.QA-FORMAT)" section spends ~450 words on question-asking etiquette largely orthogonal to skill creation, and the "Two validators, two roles" paragraph is dense enough to need a table. Matches anchor 3 — some unnecessary explanation that could be tightened — rather than 2 since the padding is confined to a couple of sections. | 3 / 5 |
Actionability | Copy-paste-ready commands ("deno run -A scripts/validate_skill.ts <path/to/skill-directory>", "wc -c <SKILL.md>"), a concrete directory tree, per-IDE path tables, good/bad description examples, and a self-review rubric. Matches the 5 anchor: fully executable guidance covering the common cases. | 5 / 5 |
Workflow Clarity | Six clearly sequenced phases (Discovery through Iterate) with a BLOCKING verification gate ("do NOT finish until it exits 0"), an explicit error-recovery feedback loop ("If it fails, FIX the description and re-run"), and checklists. Matches the 5 anchor exactly. | 5 / 5 |
Progressive Disclosure | References (references/workflows.md, references/output-patterns.md) are real, one level deep, and clearly signaled with when-to-read labels, and scripts are wired into the phase workflows. Falls short of 5 because the body sits at ~19.5k chars against its own stated 20k-char limit, with the Control Primitives Map table and the Priming section as obvious candidates for relocation to references/. | 4 / 5 |
Total | 17 / 20 Passed |