Passed 12 of 23 test cases with the skill, and 6 of 23 without it.
Without skill25%
With skill54%
Checks it passed that the model alone missed
Ends the review with CONCERN.
Evaluates script rerun idempotency.
Ends the review decision with BLOCK.
and 6 more in the full report
9 fixed4 passed either way9 still failing2 broke
Pass-rate lift
+29 pp
Tokens
+8%
Agent turns
0%
Response time
-10%
Model gemini-3.6-flashMeasured 9/28/2026v413 of 24 cases passed with skill · 23 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 9 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 9
Prompt
Review this workflow step: `run: echo "Issue title: ${{ github.event.issue.title }}" > title.txt`. The developer argues issue titles only contain plain text. Evaluate command injection risk and state your decision.
What was checked, without and with the skill
Checked
Without skill
With skill
Warns that inline expression expansion of untrusted inputs allows command injection.
Missed
Passed
Ends the review decision with BLOCK.
Missed
Passed
case-21 · gemini-3.6-flash · 9/28/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md18 linessingle file
GitHub Action Review
Read:
.agents/agents/ci-integration-reviewer.md
.agents/rules/testing.md
Review the actual changed scripts, jq filters, action definition, and tests.
Review fallow's MCP server for tool contract clarity, output stability, and agent usability. Use when changes touch crates/mcp or AI-agent integration surfaces.