Content
52%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 curated workflow (detect, review, install, verify) is genuinely actionable with executable curl examples and a verification step, but it is buried under a large, irrelevant, auto-generated CLI reference that wastes tokens and adds duplicate and even truncated entries. The skill needs aggressive pruning of the generated subcommand dump and proper externalization of reference material.
Suggestions
Remove or externalize the auto-generated omniroute subcommand reference (autostart, config, env, setup, update) — it is unrelated to the skill's stated purpose of collecting agent skills and consumes most of the token budget.
Deduplicate the repeated 'config get'/'config set'/'config list' sections and fix the truncated 'Generate OpenCode config (alias for' sentence.
Add error-recovery guidance for the workflow (auth failures, empty matchedSkills, failed installs) to strengthen the feedback loop around the install step.
If CLI reference material is needed, move it to a references/ file that actually exists and signal it clearly from SKILL.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Roughly 250 lines of the body are an auto-generated OmniRoute CLI subcommand reference (autostart, config, env, setup, update) unrelated to the skill's stated purpose, including duplicate 'config get'/'config set'/'config list' entries and a truncated sentence ('Generate OpenCode config (alias for'). The curated workflow section is efficient, so this sits above anchor 1 but is still noticeably padded with irrelevant material. | 2 / 5 |
Actionability | The curated section provides copy-paste-ready curl commands with auth headers, JSON request bodies, an example JSON response, per-tool install paths, and verification ls commands. Minor gaps: the auto-generated subcommand docs list flags with no argument descriptions, and the quick-start script silently does nothing when no tools/skills are found. | 4 / 5 |
Workflow Clarity | Steps 1-4 are clearly sequenced (detect → review matched skills → install → verify) with a relevance-score table, an explicit verification step, and a full quick-start script. Missing error-recovery guidance (e.g. what to do on auth failure or empty matchedSkills) keeps it below anchor 5. | 4 / 5 |
Progressive Disclosure | No bundle files exist at all, and the ~250-line CLI reference is content that clearly belongs in a separate file yet is inlined in the body. The only external pointer ('docs/frameworks/SKILLS.md') is not part of any bundle, leaving the structure minimally organized around the skill's actual purpose. | 2 / 5 |
Total | 12 / 20 Passed |