Conciseness

The body is lean and assumes competence: "Find the sudden change before it burns a budget" followed immediately by tool calls (`pull_google_ads_performance`, `pull_meta_ads_performance`, ...) and domain-specific heuristics like "a tracking outage reads as a conversion cliff; a new campaign launch reads as a spend spike; a weekend reads as a volume drop". No padding or explanation of concepts Claude already knows, matching the 5 anchor 'Lean and efficient; assumes Claude's competence; every token earns its place'.

Actionability

Concrete tool names are given (`list_connected_accounts`, five `pull_*_ads_performance` variants) and the action section is specific ("Recommend alert thresholds... Do not claim alerts were installed", "pause, budget cut, creative swap", route to **pre-pause-analysis**). It sits at 'Mostly executable guidance; concrete code or commands with minor gaps' because the threshold logic is underspecified — "set sane alert thresholds" never defines what a sane threshold is (window, deviation multiple), and the baseline window to pull is left open.

Workflow Clarity

The numbered sequence (baseline → outliers → diagnose → act) is clear with explicit checkpoints: "State the period and missing data before calling a change abnormal", "Confirm the cause before recommending action", and the approval gate "setting alerts or pausing is a change and waits for approval". This matches 'Clear sequence with most checkpoints present; minor validation gaps' — validation before acting is present (so the destructive/batch cap does not apply), but there is no post-action re-check loop (e.g., verifying thresholds fire correctly after configuration), keeping it below the 5 anchor.

Progressive Disclosure

The skill is under 50 lines with no bundle files and no need for external references; per the rubric's simple-skill guideline, well-organized sections alone warrant a 5. Sections are clearly headed (numbered workflow steps plus a Rules section), and the single cross-reference to **pre-pause-analysis** is clearly signaled rather than buried.

Specificity

The description lists multiple concrete, domain-specific actions: "Catch spend spikes, CTR or conversion-rate drops, budget runaway, and metric outliers across campaigns, and set sane alert thresholds" — four named anomaly types plus threshold setting. This matches the anchor 'Lists multiple specific concrete actions; comprehensive coverage' (compare the 5-anchor PDF example with 4 actions); it is above the 4 anchor because coverage of the ad-anomaly domain is complete rather than having minor gaps.

Completeness

Both parts are explicit: the what is "Catch spend spikes, CTR or conversion-rate drops, budget runaway, and metric outliers across campaigns, and set sane alert thresholds", and the when is "Use when a user asks why spend or performance changed suddenly, wants monitoring or alerts, or something looks off in the numbers". This directly matches the 5 anchor 'Clearly and explicitly answers both what AND when with concrete trigger phrases'; the 4 anchor would require a less explicit when-clause.

Trigger Term Quality

Natural user phrasing is present: "why spend or performance changed suddenly", "wants monitoring or alerts", "something looks off in the numbers", plus terms like spend spikes, CTR, conversion-rate drops, budget runaway. This fits 'Good keyword coverage; a few natural terms missing' — common synonyms such as overspend/overspending, pacing, ROAS, or CPA/CPM (which the skill body actually covers) are absent, keeping it below the comprehensive 5 anchor.

Distinctiveness Conflict Risk

The ad-campaign anomaly niche is clear and mostly distinct, but the trigger phrases "wants monitoring or alerts" and "something looks off in the numbers" are not domain-qualified and could overlap with generic monitoring or non-ad reporting skills. This fits 'Mostly distinct; minor overlap risk with closely related skills' rather than the 5 anchor's 'minimal conflict risk'.