Content
82%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, highly actionable instruction skill with real one-level-deep references and executable script commands. The main gaps are some inlined general marketing knowledge that could be trimmed, implicit rather than explicit error-recovery loops, and an orphaned reference file (copy-frameworks.md) whose content is duplicated inline instead of being linked.
Suggestions
Link references/copy-frameworks.md from the 'Ad Copy Frameworks' section (e.g., 'For the full framework selection matrix by product type, see copy-frameworks.md') and trim the inlined PAS/BAB definitions to a one-line summary, since the detailed templates already live in ad-copy-templates.md.
Make the weekly review workflow's feedback loops explicit: state what to do when a check fails or a metric misses target (e.g., 'if CPA > target for 2 consecutive weeks, run the optimization levers in order and re-check'), rather than leaving the troubleshooting ladders as a separate lookup table.
Tighten sections that restate general marketing knowledge Claude already has (e.g., 'Human faces (real, not stock)', 'Captions always (85% watch without sound)', generic PAS/BAB explanations), keeping only the platform-specific or counterintuitive heuristics that earn their tokens.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is efficient — dense tables, checklists, and bullets with no padded prose — but it restates marketing knowledge Claude already has (PAS/BAB formula definitions, generic creative tips like 'Human faces (real, not stock)' and 'Captions always'), which could be trimmed or delegated to the references. | 4 / 5 |
Actionability | Guidance is fully executable: exact script invocations with arguments ('python3 scripts/roas_calculator.py --spend 5000 --revenue 18000 --conversions 120 --clicks 2400 --margin 0.7'), a concrete naming convention with examples and explicit placeholder handling ('use the current year/quarter — {YYYY}/{Qn} are placeholders'), ordered optimization ladders for high CPA/low CTR/high CPM, and a checkbox pre-launch checklist. | 5 / 5 |
Workflow Clarity | The lifecycle is clearly sequenced (context gathering → platform selection → structure → creative → optimization → reporting → pre-launch validation), with checkpoints like 'Conversion tracking tested with real conversion' and a weekly review loop ('Run both tools on the week's numbers, then review'). Feedback loops (what to do when a check fails) are implied by the troubleshooting ladders rather than explicitly wired into the workflow, keeping it just below the top anchor. | 4 / 5 |
Progressive Disclosure | Four reference files and two scripts are all real, one level deep, and clearly signaled with purpose ('See references/audience-targeting.md' for detailed targeting strategies; scoring-system.md linked from the tools table). However, references/copy-frameworks.md exists in the bundle but is never referenced from SKILL.md — its framework selection matrix is undiscoverable — and the 'Ad Copy Frameworks' section inlines the same PAS/BAB formulas that the reference files already cover, so content placement is not fully optimal. | 4 / 5 |
Total | 17 / 20 Passed |