Content
92%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.
An exemplary router/index skill: lean routing table with per-task file guidance, concrete verification commands for every version-sensitive fact, and one-level-deep references that all resolve. The only real cost is the dense dated version-snapshot paragraph, which is time-sensitive information sitting outside any 'old patterns' framing.
Suggestions
Move the version snapshot into its own short section (e.g. '## Audit anchor') or trim it to gem names plus the lookup commands, so the dated numbers read as a checkpoint rather than body content.
Consider dropping the cross-referenced declarative-config parenthetical ("targets schema v1.0.0-rc.3; owner: ...") from the snapshot paragraph — it belongs in that sibling skill, and it adds tokens here without changing what the agent does next.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean — a routing table, a stability caveat, lookup commands, and cross-references with no explanation of concepts Claude already knows. The "Released snapshot verified 2026-10-25" paragraph packs ~15 specific version numbers and a date; it is mitigated ("Treat this as an audit anchor, not a version-alignment rule") but is the one block of time-sensitive detail that could be trimmed or relocated. Not 5 because of that dense version block; not 3 because everything else earns its tokens. | 4 / 5 |
Actionability | Concrete copy-paste commands are given for every fact ("gh api repos/open-telemetry/opentelemetry-ruby/releases --paginate", "gem specification <gem> dependencies --remote --yaml"), and the reference table maps each task to a specific file. Not 4 because the guidance is fully executable — routing decisions and fact lookups both have exact commands or file targets. | 5 / 5 |
Workflow Clarity | The single action is unambiguous and stated up front ("Load only the reference that matches the task; each reference is self-contained"), and the table makes reference selection deterministic; version questions are routed to the lookup commands with the snapshot framed as an audit anchor. Not 4 because there are no gaps or ambiguity in what to do first, and no risky operations lack validation. | 5 / 5 |
Progressive Disclosure | A clear overview with a well-signaled table of one-level-deep references (all five files verified to exist, and none of them nest further references), with the 453 lines of detail appropriately split out of SKILL.md. Not 4 because navigation is fully deterministic via the "Use when" column and links; cross-skill pointers are clearly labeled as such. | 5 / 5 |
Total | 19 / 20 Passed |