Content
67%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 a well-structured, highly actionable router spec: a complete signal→route table, a branched workflow, and concrete output templates. Its main weakness is token efficiency — the marketing-flavored justification section and redundant restatements of the routing table dilute otherwise lean content.
Suggestions
Cut or compress the 'Why This Is the Killer Command' section — phrases like 'killer command' and 'cognitive tax' are promotional justification that adds no routing instruction; keep at most one sentence on why the router exists.
Trim the intro and Examples section, which repeat routing outcomes ('gross retention dropped → CCO') already encoded in the table; keep 2-3 diverse examples and drop those that duplicate table rows.
Specify the mechanics for 'invoke that cs-* agent directly' and 'Log the routing decision via decision-logger' with the exact command or call form, which would also lift actionability toward anchor 5.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The signal→route table and workflow are dense, plugin-specific facts that earn their tokens, but "Why This Is the Killer Command" ("the killer command", "cognitive tax", "This is also where persistent memory pays off") is promotional padding Claude doesn't need, and the intro plus Examples section restate routing outcomes the table already encodes. Fits anchor 3 (mostly efficient, some unnecessary explanation) better than anchor 4, which has only minor trims. | 3 / 5 |
Actionability | Concrete keyword→agent mapping ("burn, runway, fundraise... | cs-cfo-advisor"), named commands (/c-level-agents:brief, /c-level-agents:office-hours, decision-logger), and exact output templates for all three route types make the guidance executable. Falls short of anchor 5 because the invocation mechanics ("invoke that cs-* agent directly", "Log the routing decision (raw layer) via decision-logger") are stated as direction without the specific command form. | 4 / 5 |
Workflow Clarity | The five-step workflow cleanly sequences single-role, multi-role, and ambiguous branches with a designated fallback (office-hours) — anchor 4's clear sequence. No validation checkpoints exist, but the operation is non-destructive routing so the destructive/batch cap doesn't apply; it misses anchor 5 on lacking any feedback loop (e.g., tie-breaking when one question hits keywords in two rows of a single role, or what to do if an agent is unavailable). | 4 / 5 |
Progressive Disclosure | Sections are well-organized and all routing content is appropriately inline for a router skill (no bundle files exist to split into), with external links (cs-chief-of-staff agent, chief-of-staff and context-engine skills) clearly signaled one level deep in the Related section. Anchor 4 fits: good structure with minor organization gaps — the Output examples are long enough they could live in a reference file if the bundle grows. | 4 / 5 |
Total | 15 / 20 Passed |