Content
92%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 model of lean, instruction-only skill writing: one concrete tool call, a clear precedence rule, and specific guidance for using the business's voice and facts without fabrication. The only minor gap is the absence of brief examples showing the desired output style.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~9-line body is lean with zero padding: "call the andrewrendon connector's `get_business` tool once in the conversation, unless you already have its result" plus three tight rules. It assumes Claude's competence and explains nothing the model already knows, matching the 'every token earns its place' anchor. | 5 / 5 |
Actionability | The guidance is concrete and executable for an instruction-only skill: a named tool (`get_business` on the andrewrendon connector), an explicit call-once condition, and specific behavioral rules including an edge case ("say that it isn't written down rather than inventing it"). It falls short of 5 only because no examples illustrate the desired voice, signing, or phrasing of the missing-facts disclaimer. | 4 / 5 |
Workflow Clarity | For this simple two-phase skill the sequence is unambiguous: fetch context first (with an idempotency checkpoint "unless you already have its result"), then apply the three rules, with an explicit fallback when notes lack needed facts. No destructive or batch operations are involved, so the simple-skill exception applies and the single action is fully clear. | 5 / 5 |
Progressive Disclosure | The skill is under 50 lines with no need for external references (the business details live in the connector, not in bundle files), and the body is well organized as an instruction line plus a bulleted rule list. Per the rubric's simple-skill guidance, this earns a 5 with no separate files required. | 5 / 5 |
Total | 19 / 20 Passed |