Content
75%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.
A dense, disciplined procedural skill: correctly sequenced with real validation loops, literal output templates, and conditional external references. Its main cost is compression — multi-clause mega-sentences hide critical requirements — and its references are unverifiable from this bundle since they live in the parent skill's directory.
Suggestions
Break the ~90-word sentences in steps 1 and 3 into short labeled lists (e.g. the flow-header fields, the per-message checklist) so no requirement is buried mid-sentence.
Include one condensed worked example of a finished flow entry (status line + one message) inline so the quality bar does not depend entirely on the external 'worked-flows.md'.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body explains nothing Claude already knows and has no padding — every line is procedural. It sits between anchors 4 and 5: efficient overall, but steps 1 and 3 are ~90-word single sentences that bury critical requirements (the flow-header field list, the per-message checklist) and could be structured with no token cost. Not anchor 5's 'every token earns its place' in effective form, clearly above anchor 3. | 4 / 5 |
Actionability | Highly concrete for an instruction-only skill: exact paths ('lifecycle/flows/<flow-slug>.md'), literal templates ('not ready: answer Q3, Q5', the 'Graphic needed:' line format), and a worked hypothesis example ('abandoned checkout: ≥ 4% of recipients order within 7 days'). It falls short of anchor 5 because the core deliverable's syntax is delegated to the external 'loadable-flow-format.md' and no worked output example appears inline — minor gaps, so anchor 4. | 4 / 5 |
Workflow Clarity | Steps 1-7 are clearly sequenced (wire, plan, write, self-check, gaps, hypothesis, art), step 4 is an explicit validate-and-rewrite loop, and step 5 gates output status. This meets much of anchor 5, but the extreme sentence density makes some checkpoints implicit (e.g. 'Check plan reality' criteria) and final verification is delegated to the parent skill — between anchors 4 and 5. | 4 / 5 |
Progressive Disclosure | References are one level deep and well-signaled with purpose and conditional loading ('sms-push-whatsapp.md when the flow has those legs', 'worked-flows.md (quality bar)'), matching anchor 4's 'references mostly clear'. It is not anchor 5 because all references live outside this skill's own bundle ('../email-lifecycle/references/'), none are present here to verify, and the body inlines heavy procedural detail that makes the SKILL.md more than a lean overview. | 4 / 5 |
Total | 16 / 20 Passed |