Content
78%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, professional instruction skill: a clear skill contract, seven sequenced steps with NEEDS_INPUT checkpoints and a 'Done when' checklist, concrete segment examples, and clean one-level-deep reference pointers. It loses points on repetition of the campaign-architect handoff and on delegating the funnel-stage map and handoff-summary shapes entirely to external files without a minimal inline example. Workflow validation exists but lacks a recovery loop for wholly unusable exports.
Suggestions
Add a compact inline example of the funnel-stage targeting map and the handoff-summary shape (a few lines each), so the two central output artifacts are concrete without chasing external references (actionability).
State the campaign-architect handoff once (e.g. in Next Best Skill) and cut the repetitions in the intro, Skill Contract, scope guard, and Reference Materials (conciseness).
Add an explicit recovery step for wholly unusable exports — what to ask the user for and what minimal artifact to return when parsing fails entirely, complementing the existing no-value-columns fallback (workflow_clarity).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and assumes Claude's competence (no tutorials on what CRM or lookalike audiences are), but the campaign-architect handoff is repeated five times (intro, Skill Contract, scope guard, Reference Materials, Next Best Skill) and the opening paragraph restates the description nearly verbatim. This fits 'efficient; minor instances of over-explanation that could be trimmed' rather than the lean score-5 anchor. | 4 / 5 |
Actionability | Guidance is concrete for an instruction-only skill: named segment examples ('repeat-buyers-90d', 'high-AOV', 'pricing-page-visitors'), explicit suppression windows (14–30 days), typed profile weights (0.15/0.30/0.10), and an exact save-path template. It stops short of score 5 because the two central output artifacts — the funnel-stage map and the handoff summary — are delegated to external references without even a compact inline example of their shape. | 4 / 5 |
Workflow Clarity | Seven clearly sequenced steps with a 'Done when' completion checklist, NEEDS_INPUT flagging for missing columns, and an explicit degraded-input fallback in step 2 ('still return the segment plan... flag only the missing fields NEEDS_INPUT'). Not score 5: there is no step-level feedback loop, e.g. what to do when an export fails to parse at all or when segment sizes are unusable, so checkpoints are present but not complete. | 4 / 5 |
Progressive Disclosure | No bundle files exist under ./references/, ./scripts/, or ./assets/, and the body is a ~65-line overview that keeps bulk detail in clearly signaled one-level-deep references, each annotated with its purpose ('roas-benchmark.md — ROAS framework, A-dimension items, typed profiles'). This matches the score-5 anchor: clear overview, well-signaled references, content appropriately split. | 5 / 5 |
Total | 17 / 20 Passed |