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 dense but effective routing guide: an explicit decision checklist, concrete command forms, and specific failure-handling guidance make it highly actionable. Its main weaknesses are repetition of chain/silence rules across the guardrails and the unspecified $TARGET syntax, plus a monolithic guardrail block that would navigate better as a reference file.
Suggestions
Consolidate the overlapping `--chain`/`--silence` guardrail bullets into a single decision table keyed by (dependency? result needed?) to remove repetition and shrink token cost.
Specify the exact $TARGET syntax (agent identifier format) next to the heredoc examples so the command is fully copy-paste ready.
Move the long edge-case guardrails (nested-ask lineage, followup correction channels, diagnostics-only commands) into a references file and keep only the top routing rules in SKILL.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | There is no padding or explanation of concepts Claude already knows, but the Guardrails section restates `--chain`/`--silence` routing across many overlapping bullets (e.g., chain rules appear in items 2, and at least five guardrail bullets), which could be consolidated. This matches 'Mostly efficient but includes some unnecessary explanation or could be tightened' rather than level 4's 'minor instances of over-explanation'. | 3 / 5 |
Actionability | The heredoc command templates are executable, flag semantics are concretely defined per case, and error recovery is specific ('If CCB says `ask --chain requires an active parent job`, retry once with plain ask'). Not level 5 because $TARGET's format (agent name, ID, address?) is never specified, leaving a gap in copy-paste readiness for the core command. | 4 / 5 |
Workflow Clarity | The Decision Card gives a clear, ordered pre-send checklist (delegation need → dependency gate → result intent → request fidelity), followed by the exact command form and explicit post-send behavior ('end the turn... do not poll'). Not level 5 because the dense interleaving of routing rules between Decision Card and Guardrails leaves minor ambiguity about which rule wins in edge cases, and there is no step verifying the submission was accepted. | 4 / 5 |
Progressive Disclosure | No bundle files exist, and the body is well organized into clearly signaled sections (Decision Card, Guardrails, command forms, post-send rules) that fit a single-purpose routing skill. Not level 5 because the ~85-line body exceeds the under-50-line simple-skill exception and the 25-bullet Guardrails section is a dense block that could be split into a reference file for easier navigation. | 4 / 5 |
Total | 15 / 20 Passed |