code-review-adversarial
Adversarial code review of spec/quality with a single agent
- License 许可
- MIT
- 最近更新
- 2026-11-12
- Stars
- ★ 21
- Forks
- 1
01 前言 Front Matter
- name
- code-review-adversarial
- description
- Adversarial code review of spec/quality with a single agent
03 SKILL.md 正文
Perform an adversarial /code-review using instructions supplied to this skill.
A reviewer is not limited to just reading the code but can also probe it by executing it.
There is one instruction to override what is in the /code-review skill: the usage of parallel sub-agents.
First classify the review context as large (> 100k tokens) or small based on the review context (spec and changeset).
You can determine this by making a very rough estimate with a word counting tool- do not read any information into your context.
For a small review context, do not use 2 parallel sub-agents instead have one sub-agent do a 2-pass review- the first for spec, and the 2nd for standards.
04 同仓库更多 Skills
easy-issues
Find and work through easy issue-tracker tasks autonomously, simplest first, using one fresh subagent per issue in sequence. Use for clearing straightforward bugs, documentation fixes, and tiny features without operator help.
wrap-up
After implementing, commit and finish your work
verify
Verify software by intentional testing.
update-working-copy
Fetch the latest from the remote. Update the working branch or switch to main if the branch is defunct.
tickets
Break a plan, spec, or conversation into tracer-bullet issues for this repo, creating a parent tracking issue with sub-issues when publishing multiple tickets.
spec
Turn the current conversation into a spec and publish it to the project issue tracker: no interview, just synthesis of what you've already discussed.