Content
86%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 well-structured, lean analytical skill that delegates detail to one-level-deep references and provides highly actionable thresholds, formulas, and tool names. The only notable gap is the absence of an explicit linear pipeline with error-recovery feedback loops, though validation checkpoints for mutations are present.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean, assumes Claude knows paid-social basics (no definitions of ROAS/ad sets), and every section earns its place; the inline account-baseline JSON schema and tool-surface listing are slightly long but justified, keeping it just below a 5. | 4 / 5 |
Actionability | Concrete decision rules with numeric thresholds ("frequency > 3.0 with rising CPM", "50-events-in-7-days", ">30% anomaly"), an explicit update formula (`rolling30d = (0.7 × previous_rolling30d) + (0.3 × recent7d × (30/7))`), and exact tool names (`runScript`, `ads.graphParallel`, `pauseAdSet`) make the instruction-only guidance highly executable. | 5 / 5 |
Workflow Clarity | Operating principles #1 ("Confirm before writing") and #4 ("Recommend, then act — wait for approval before mutating") provide explicit validation checkpoints for mutation operations, and the reference-framework table plus conditional handoffs sequence the workflow; it lacks a single strict numbered pipeline with error-recovery loops, so it stops at 4. | 4 / 5 |
Progressive Disclosure | A clear "Reference framework — when to read what" table routes user intent to one-level-deep, real bundle files (analysis-heuristics.md, creative-fatigue.md, learning-phase.md, audience-strategy.md, industry-benchmarks.md, campaign-structure-guide.md), keeping the SKILL.md an overview with well-signaled, easy navigation. | 5 / 5 |
Total | 18 / 20 Passed |