Passed 17 of 20 test cases with the skill, and 6 of 20 without it.
Without skill32%
With skill86%
Checks it passed that the model alone missed
Reports a packaging error.
Requires reporting a packaging error.
Refrains from returning mock JSON budget updates.
and 17 more in the full report
14 fixed5 passed either way1 still failing2 broke
Pass-rate lift
+55 pp
Tokens
-36%
Agent turns
0%
Response time
-12%
Measured 9/19/2026v4 · latest is v619 of 22 cases passed with skill · 20 comparable
Tokens, turns and time describe resource use, not a dollar price.
This benchmark measured an earlier version of the skill.
A newer attempt on gemini-3.8-flash on 10/18/2026 did not produce a publishable comparison (not comparable). The snapshot above is the previously measured run.
See it in action
One of 12 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 12
Prompt
When resolving the location of the active workflow file, should the file system path include the 'skills/' subfolder segment or locate 'meta-ads/manage/SKILL.md' directly under the root? Specify the correct path.
What was checked, without and with the skill
Checked
Without skill
With skill
Identifies <plugin-root>/meta-ads/manage/SKILL.md as the canonical path.
Missed
Passed
Rejects the <plugin-root>/skills/meta-ads/manage/SKILL.md path.
Missed
Passed
case-07 · gemini-3.6-flash · 9/19/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md3 linessingle file
Canonical NotFair workflow
Read ../../meta-ads/manage/WORKFLOW.md completely, then follow it as the active workflow. Normalize that path from the directory containing this wrapper: the canonical file is <plugin-root>/meta-ads/manage/WORKFLOW.md, not <plugin-root>/skills/meta-ads/manage/WORKFLOW.md. Resolve every relative reference from the canonical file against <plugin-root>/meta-ads/manage/. If the canonical file cannot be read, stop and report the packaging error; never substitute a similarly named skill from another plugin.