Use when reviewing a pull request, diff, branch, or code change for bugs, regressions, missing tests, and release risk.
SkillScore68Maintained
Automatic
◆Coding pack
Last verified 8/3/2026
What it changes
Passed 14 of 20 test cases with the skill, and 11 of 20 without it.
Without skill55%
With skill68%
Checks it passed that the model alone missed
The review excludes style-only comments.
The output uses templates/review-output.md.
The response explicitly mentions residual test risk.
and 5 more in the full report
7 fixed8 passed either way3 still failing4 broke
Pass-rate lift
+14 pp
Tokens
+45%
Agent turns
0%
Response time
+33%
Model gemini-3.6-flashMeasured 8/3/2026v115 of 22 cases passed with skill · 20 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 7 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 7
Prompt
Review PR #304 in auth-service which fixes a minor typo in a log message across logger.py lines 15-20. Assume all existing tests pass and no functional code was changed.
What was checked, without and with the skill
Checked
Without skill
With skill
The response explicitly states that there are no issues.
Missed
Passed
The response explicitly mentions residual test risk.
Missed
Passed
case-04 · gemini-3.6-flash · 8/3/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md30 lines3 files
PR Review Workflow
Goal
Find the highest-impact issues in the change. The output is for an engineer deciding whether the PR can merge.
Process
Read the diff and the surrounding code needed to understand behavior.
Identify changed contracts, data flow, error handling, user-visible behavior, and tests.
List findings first, ordered by severity.
Keep summaries brief.
Finding Rules
A finding must describe a concrete risk.
A finding must include file and line evidence.
Do not include speculative issues without stating the assumption.
Do not include style-only comments.
If there are no issues, say that clearly and mention residual test risk.
Output Format
Use templates/review-output.md.
End with templates/completion-receipt.md. The receipt must separate checked claims from unverified claims.
Ambient Memori long-term memory for Claude Code via local Bash. MUST TRIGGER on essentially every non-trivial user turn: run recall before drafting any substantive response, whether or not the user mentioned memory, a prior session, or past work. The default is to check memory first; do not wait to be asked. Always fire before external lookups (WebSearch/WebFetch) and before answering any question about preferences, prior decisions, constraints, project history, status, past work, or any non-tri