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 tight, high-signal instruction skill: prescriptive heuristics, named tools, and a rules section that reinforces pre-launch checks, organized as a clear five-step workflow. The gaps are minor — a couple of high-level directions without a method or source, and no explicit underperformance recovery loop.
Suggestions
Make 'map your vertical to LinkedIn's taxonomy' actionable: point to where the taxonomy lives (platform UI or an export) or name the tool/UI step that performs the mapping.
Name the mechanism for benchmark comparison in section 3 — e.g. which tool pulls the account's historical CPC/CPM/CPL — instead of just 'compare with the account's recent history'.
Add one recovery line to section 5 for the common failure case: what to check and change when delivery or CPL is underperforming after launch.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean throughout: every line is a LinkedIn-specific heuristic ('100,000-300,000 is the sweet spot', '3 pre-filled fields converts roughly 12-15%'), a named tool call, or a rule — no padding and no explanation of concepts Claude already knows. Nothing trimmable was found, so it matches the 'every token earns its place' anchor. | 5 / 5 |
Actionability | Mostly concrete and executable: named tools ('list_connected_accounts', 'list_audiences', 'pull_linkedin_ads_performance'), numeric thresholds ('below ~50,000 members delivery suffers'), and specific tiering and form-field guidance. A few spots stay high-level — 'map your vertical to LinkedIn's taxonomy' gives no mapping or source, and 'Compare CPC, CPM, CPL, and CTR with the account's recent history' names no tool or method — matching the 'minor gaps' anchor rather than fully executable. | 4 / 5 |
Workflow Clarity | Clear numbered sequence (audience → ABM → format → forms → 'Verify and iterate') with explicit checkpoints ('Confirm campaign state with list_campaigns', 'Check estimated audience size before launch'). Falls short of the top anchor because there is no error-recovery loop for underperforming delivery — post-launch iteration is only implied by 'Configure alerts and new expansion audiences'. | 4 / 5 |
Progressive Disclosure | A short (~55-line) single-file skill with no bundle files and nothing inlined that belongs in a separate file; sections are numbered and well-organized, and cross-skill pointers ('audience', 'attribution', 'media-plan', 'ad-copy-generation', 'platform-linkedin') are clearly signaled, satisfying the simple-skill guideline. | 5 / 5 |
Total | 18 / 20 Passed |