Content
81%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 content delivers a clear, sequenced decision procedure with concrete command templates and explicit error-recovery guidance, and it avoids padding almost entirely. Its weaknesses are minor: some redundancy across guardrail bullets, missing literal examples for the silence/compact variants, and a single-file structure that, while well-organized, is denser than the ideal progressive-disclosure shape.
Suggestions
Consolidate the repeated chain/silence routing rules (Guardrails bullets around lines 49–52 overlap the Decision Card's dependency gate) into one canonical statement to cut redundant tokens.
Add a fully literal example invocation (real target name and short message) and templates for the `--silence` and `--compact` variants so all common cases are copy-paste ready.
Consider moving the long exception-handling guardrails (followup injection states, lineage routing, diagnostics) into a references/ file, keeping SKILL.md as a lean overview with one-level-deep pointers.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and prescriptive with no explanation of concepts Claude already knows, but the chain/silence routing rules recur across several guardrail bullets (e.g., "Use `--chain` only for a real child dependency; use `--silence` for independent no-result work" restates the Decision Card's dependency gate), so minor tightening is possible. It sits between 'mostly efficient' (3) and 'lean, every token earns its place' (5) — efficient with a few redundant restatements, matching anchor 4. | 4 / 5 |
Actionability | Concrete heredoc command templates cover the plain and full-flag cases, and error handling is explicit ("If CCB says `ask --chain requires an active parent job`, retry once with plain `ask`"). Not 5 because no fully literal example with a real target and message is given and the `--silence`/`--compact` variants have no template; not 3 because the guidance is executable rather than pseudocode. | 4 / 5 |
Workflow Clarity | The numbered Decision Card (need delegation → dependency gate → result intent → request fidelity) gives a clear sequenced procedure, followed by submit → stop → report with explicit error-recovery loops (chain-error retry, permission-denied reporting). This is messaging/delegation, not a destructive or batch operation, so the validation cap does not apply; the sequence plus feedback loops matches anchor 5. | 5 / 5 |
Progressive Disclosure | The skill has no bundle files, references no external paths, and is organized into clean sections (Decision Card, Guardrails, templates) with everything appropriately inline. Not 5 because the body runs ~87 lines — above the under-50-line simple-skill threshold under which well-organized sections alone would earn a 5 — and the dense guardrail set could arguably live in a reference file; not 3 because structure is good and nothing that belongs elsewhere is misplaced. | 4 / 5 |
Total | 17 / 20 Passed |