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.
The body is an exceptionally actionable, well-sequenced guide with concrete per-IDE formats, templates, and a bundled validation script — workflow clarity and actionability are top-tier. Its main weakness is the inlined, densely packed Question Format section, which hurts both conciseness and progressive disclosure.
Suggestions
Move the Question Format (FR-UNIVERSAL.QA-FORMAT) policy into a reference file such as references/question-format.md and keep a two-line summary plus a pointer in SKILL.md — this addresses both the conciseness and progressive-disclosure gaps.
Split the single ~300-word bullet on option wording into a short checklist of discrete rules (labels, group pointers, coined terms, the self-containment test) so each rule is scannable instead of one paragraph.
Signal references/workflows.md and references/output-patterns.md where they are used (Step 2, when planning the body) in addition to Step 3's writing rules, so navigation to the bundled detail is discoverable at the point of need.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Most sections (paths table, creation process, validation) are lean and carry only project-specific knowledge Claude would not know, but the "Question Format (FR-UNIVERSAL.QA-FORMAT)" section is a padded wall of text — the single bullet at line 20 packs ~300 words of general option-wording policy into one paragraph, much of it generic question-asking guidance rather than command-authoring specifics. It fits 'mostly efficient but includes some unnecessary explanation or could be tightened' better than the verbose-2 anchor, since the rest of the body is genuinely dense with non-obvious detail. | 3 / 5 |
Actionability | The body is fully executable throughout: exact per-IDE paths in a table (`.claude/commands/<name>.md`, `~/.config/opencode/commands/<name>.md`), concrete frontmatter keys per IDE, copy-paste-ready file templates for all three IDEs, a YAML-quoting rule with a concrete failure case, and a runnable validation command (`deno run -A scripts/validate_command.ts <path> --ide <claude|cursor|opencode>`) whose script exists in the bundle. Not the level below because nothing is pseudocode or left as a high-level hint. | 5 / 5 |
Workflow Clarity | The five-step creation process (collect invocations → plan → write → validate → iterate) is clearly sequenced with an explicit validation checkpoint and a feedback loop ("Fix every error it reports and run it again"), plus a post-use iteration step. This matches the anchor with explicit validation steps and error-recovery loops; no destructive/batch cap applies since the skill only writes one file. | 5 / 5 |
Progressive Disclosure | References are real, one level deep, and clearly signaled ("see `references/workflows.md` and `references/output-patterns.md`", plus `scripts/validate_command.ts`), which is 4-anchor territory — but the ~10-line, ~350-word Question Format policy is inlined monolith content that clearly belongs in its own reference file, pulling the dimension back to 'some structure but content that should be separate is inline'. It does not reach 2 because the rest of the body is well-sectioned and nothing is deeply nested. | 3 / 5 |
Total | 16 / 20 Passed |