Content
88%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 operational contract: lean per-step instructions backed by a verified 770-line template file, an explicit evidence gate with a NEEDS_INPUT feedback loop, and a done-when checklist. Weaknesses are minor — repeated STAR-ownership disclaimers across four sections and a duplicated STAR reference entry.
Suggestions
State the STAR/SQS non-ownership boundary once (e.g., keep the scope guard, drop the restatements in the intro, Reference Materials, and Next Best Skill) to recover ~8 lines of token budget.
Merge the two Reference Materials bullets that both link ../../../references/star-benchmark.md into a single entry covering both the framework and skill-ownership notes.
Consolidate the WARM-save/HOT-promotion authorization rules, which currently appear in Skill Contract, Save Results, and Next Best Skill, into one section with the other referenced once.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is procedural and assumes competence (one-line steps, no explanations of known concepts), but boundary disclaimers repeat across sections: "it computes neither the Suitability/Trust/Appeal/Return dimension scores nor the SQS" (intro), the scope guard's "The Brand Fit Score X/25 is only a community-entry assessment, not STAR S or SQS", and "which this skill does not compute" (Reference Materials); two Reference Materials bullets also link the same star-benchmark.md, and WARM/HOT authorization rules appear in three sections. Not 5: this repetition is trimmable; not 3: there is no padding or explanation of things Claude already knows — the redundancy is limited emphasis, roughly 10% of the body. | 4 / 5 |
Actionability | Instruction-only skill with fully concrete guidance: each step names its deliverable (e.g., "Generate an audience persona — ≥1 named persona with bio, day-in-the-life, goals, media consumption, and a key quote"), the mode-inference rule gives examples ("a named community, subculture, or hashtag (e.g. '#BookTok', 'van-life') → niche"), the rollout is specified ("Listen & Learn → Soft Entry → Active Engagement"), and every step's fill-in template was verified in references/templates.md — including a real Brand Fit Score X/25 scoring table (templates.md:651-660). Not 4: no gaps remain; the templates cover the common cases for both modes. | 5 / 5 |
Workflow Clarity | Clear sequence (step 1 sets mode/context, then A2–A9 or N2–N7), an explicit validation gate run before any filling ("First run the evidence gate. Unsupported fields stay TBD and produce a query plan"), a feedback loop for error recovery (insufficient evidence → return NEEDS_INPUT "plus the exact queries/data fields"), and a 3-item "Done when" checklist. Not 4: checkpoints are explicit rather than implicit; this is a non-destructive analysis skill so the destructive-operation cap does not apply. | 5 / 5 |
Progressive Disclosure | The body is an overview with a single one-level-deep bundle reference that is well-signaled and verified: references/templates.md sections (## A2–## N7) match the body's anchor links exactly, and the Reference Materials list describes each link's purpose. Not 5: minor organization gaps — two Reference Materials bullets ("STAR Benchmark" and "STAR benchmark — Skill Ownership") point to the same file, and many cross-links (../../../references/star-benchmark.md, ../influencer-discovery/SKILL.md, ../../../CONNECTORS.md) resolve outside this bundle. Not 3: structure is good, references are clearly signaled, and detail is appropriately split into the templates file. | 4 / 5 |
Total | 18 / 20 Passed |