Content
85%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a well-organized hub: a subcommand-to-reference table, a short sequenced execution workflow with verification, and a complete reference index whose targets all exist. Its main weakness is inline version-number annotations that clutter an otherwise lean document.
Suggestions
Move time-sensitive version annotations (CC 2.1.108+/2.1.120+, Codex 0.123.0/0.130.0) into a dedicated "Version notes" or "Deprecated/old patterns" section so the main flow stays lean.
Consolidate the three separate blockquote guidance notes (built-in slash discovery, Claude Code setup, Codex plugin workflows) into a single short note pointing to the authoritative docs, reducing repetition.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly lean (table + short steps + reference index), but inline time-sensitive version markers ("CC 2.1.108+", "CC 2.1.120+", "Codex 0.130.0", "0.123.0+") are not isolated in a deprecated/old-patterns section, which the rubric penalizes. | 2 / 3 |
Actionability | The Quick Reference table maps each subcommand to a concrete reference file and the Execution section names specific verification commands ("harness doctor", "bash scripts/sync-skill-mirrors.sh --check", "bash scripts/ci/check-consistency.sh"), giving copy-paste-ready routing. | 3 / 3 |
Workflow Clarity | Execution is a clear four-step sequence ending in an explicit verification checkpoint (step 4 lists doctor + consistency checks), which is appropriate for a routing skill with no destructive batch operations. | 3 / 3 |
Progressive Disclosure | A concise overview (Quick Reference table + Reference Index) points one level deep to references/*.md, all of which exist on disk (init, ci, codex, cursor, harness-mem, mirrors-agents-localize, marketplace, maintenance), with clearly signaled navigation. | 3 / 3 |
Total | 11 / 12 Passed |