Passed 17 of 22 test cases with the skill, and 7 of 22 without it.
Without skill32%
With skill77%
Checks it passed that the model alone missed
Provides projected reach and frequency estimates
Includes estimated reach and frequency projections
Reserves 10% to 15% of the budget in a contingency pool
and 11 more in the full report
10 fixed7 passed either way5 still failing0 broke
Pass-rate lift
+45 pp
Tokens
+39%
Agent turns
0%
Response time
+16%
Model gemini-3.6-flashMeasured 8/17/2026v1 · latest is v217 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
This benchmark measured an earlier version of the skill.
See it in action
One of 10 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 10
Prompt
We are planning a Q3 product launch campaign running from July 1 to September 30 with a total paid advertising budget of $150,000 across Meta Ads, Google Ads, and TikTok Ads. Our main objective is driving online sales while building initial brand awareness among US millennials aged 25-40. We have existing static banners and short-form video assets ready. Please build a full paid media strategy and plan that details our platform budget splits, weekly flight schedule and pacing targets, audience targeting matrix per channel, creative rotation timeline with refresh intervals, reach and frequency estimates, measurement tracking setup, and a contingency reserve framework for unexpected CPM shifts.
What was checked, without and with the skill
Checked
Without skill
With skill
Reserves 10% to 15% of the $150,000 budget for contingency
Passed
Passed
Includes projected reach and frequency estimates
Passed
Passed
Details tracking implementation requirements for measurement setup
Missed
Passed
case-04 · gemini-3.6-flash · 8/17/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md72 linessingle file
/digital-marketing-pro:media-plan
Purpose
Generate a holistic paid media plan that coordinates budget, channels, audiences, creative, and timing across all advertising platforms. Balances reach and efficiency objectives with practical execution constraints to produce a ready-to-implement plan with clear pacing targets and contingency protocols.
Input Required
The user must provide (or will be prompted for):
Campaign dates: Start date, end date, and any blackout periods or mandatory flight windows
Total paid media budget: Aggregate budget for the campaign period with any channel-specific floors or caps
Channels available: Platforms in consideration — Google Ads, Meta Ads, LinkedIn Ads, TikTok Ads, ChatGPT Ads, programmatic display, connected TV, native, audio, out-of-home, etc.
Target audiences with segments: Audience definitions including demographics, interests, behaviors, custom audiences, lookalikes, and retargeting pools
Creative assets available: Existing ad formats and sizes, video lengths, static variants, and any creative production timelines for new assets
Geographic targeting: Markets, regions, DMAs, or countries to target with any geo-specific budget weighting
Competitive spending intelligence: Known or estimated competitor ad spend, share of voice benchmarks, and auction pressure indicators
Historical performance by channel: Past campaign data — CPC, CPM, CPA, ROAS, conversion rates — by channel and audience segment
Seasonality factors: Demand fluctuations, industry events, holidays, promotional periods, or competitive surges that affect costs or performance
Process
Load brand context: Read ~/.claude-marketing/brands/_active-brand.json for the active slug, then load ~/.claude-marketing/brands/{slug}/profile.json. Apply brand voice, compliance rules for target markets (skills/context-engine/compliance-rules.md), and industry context. Also check for guidelines at ~/.claude-marketing/brands/{slug}/guidelines/_manifest.json — if present, load restrictions and relevant category files. Check for custom templates at ~/.claude-marketing/brands/{slug}/templates/. Check for agency SOPs at ~/.claude-marketing/sops/. If no brand exists, ask: "Set up a brand first (/digital-marketing-pro:brand-setup)?" — or proceed with defaults.
Assess channel-objective fit: Evaluate each available channel against campaign objectives using reach capability, targeting precision, cost benchmarks, creative format support, and measurement reliability
Allocate budget across channels: Distribute budget using efficiency-weighted allocation — factor in historical performance, diminishing returns curves, minimum effective spend thresholds, and strategic importance per channel
Design flight schedule: Structure campaign timing as continuous, pulsing, or flighting based on objectives, seasonality, and budget — define weekly spend levels and heavy-up periods
Build audience targeting matrix: Map each audience segment to its optimal channel(s) with targeting parameters, expected reach, estimated frequency, and overlap management between platforms
Plan creative rotation: Schedule creative variants across channels with rotation frequency, fatigue thresholds (impressions or frequency caps), A/B test windows, and refresh dates for new assets
Define measurement framework: Establish KPIs per channel, tracking requirements (pixels, UTMs, offline conversion imports), attribution model, and reporting cadence
Set contingency holdback: Reserve 10-15% of budget as contingency for opportunistic scaling, underperformance reallocation, or emerging platform opportunities — define trigger criteria for deployment
Model reach and frequency estimates: Project total reach, average frequency, and effective frequency per channel and in aggregate — flag oversaturation or underspend risks
Compile unified media plan calendar: Assemble all components into a single calendar view showing budget pacing, creative rotation, audience activation, and measurement milestones week by week
Ads inside AI answers (checked 2026-10-04)
Two surfaces need special handling in any plan. Full sourcing, tracking and feed detail live in skills/paid-advertising/ads-in-ai-answers.md; follow it rather than restating numbers from memory.
ChatGPT Ads: plan it as a test line, funded from the contingency reserve (step 8).