skills.rest
wesbragagt
wesbragagtCOMMUNITY
Explore all skills by this author →
Website𝕏@wesbragagtGitHub

adreview

Runs an adversarial code review dialogue with a sibling agent over tmux.

1|Updated May 4, 2026
Search Tags:#code review#multi-agent#tmux#workmux#adversarial review#findings verification
One-click install
npx skills add https://github.com/wesbragagt/nixos-config --skill adreview-wesbragagt
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: adreview
Source: https://github.com/wesbragagt/nixos-config/tree/main/home/skills/adreview
Command: npx skills add https://github.com/wesbragagt/nixos-config --skill adreview-wesbragagt

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Single-agent code reviews can miss risks because the same model that wrote the code evaluates it. This Skill sends the current worktree to a second, independent agent in the same workmux session and manages a structured adversarial review dialogue, so findings are challenged and verified before you act on them. ## Core Features & Use Cases - Sibling reviewer selection: Automatically resolves the current tmux session and window, then targets the opposite agent (claude or omp) as the reviewer. - Structured review protocol: Requires the reviewer to cite exact files and lines, rank findings as must-fix, should-fix, or optional, and return "No findings" when evidence is lacking. - Verified reporting: Every sibling finding is checked against the repository before being accepted, rejected, or flagged as unresolved. - Use Case: After implementing a feature in a workmux session, invoke /adreview to have the sibling agent hunt for correctness, security, and performance risks, then receive a verified findings report without any code being modified. ## Quick Start Invoke /adreview with an optional target such as a branch, diff, or set of files to start an adversarial review of the current worktree.

Frequently Asked Questions about adreview

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run an adversarial code review between two AI agents?▼

Invoke /adreview from the claude or omp window of a workmux session. The skill resolves the tmux session, targets the sibling agent, and manages a review dialogue where findings must cite exact files and lines.

How does the skill send prompts to the other agent in tmux?▼

Each message is written to a temporary file, loaded as a tmux paste buffer with tmux load-buffer, pasted into the target pane, and followed by a separate Enter key. Review text is never built through shell interpolation.

Does the adversarial review modify my code?▼

No. The sibling agent is instructed to review only, without editing files, and the orchestrating agent does not modify code unless the user separately asks for fixes.

What happens if I run /adreview outside a claude or omp window?▼

The skill reports that /adreview requires the claude or omp workmux window and stops. It never selects a window from another tmux session.

How are review findings verified before being reported?▼

Every sibling finding is checked against the repository before acceptance. The final report lists accepted findings with evidence and severity, rejected findings with reasons, and unresolved items needing user input.

Related Skills

Explore More Software Engineering Skills→
ciaran-slowciaran-slow

workmux

Manage git worktrees paired with tmux windows as isolated parallel development environments.

Community
Advanced
mnemosyne-ossmnemosyne-oss
2.3k

no-mistakes

Validate committed code changes through review, test, lint, and CI pipelines.

Official
Advanced
baphledbaphled
1

check-compliance

Run linting, formatting, security scans, and tests before committing code.

Community
Intermediate
golemcloudgolemcloud
1.5k

golem-wait-for-external-input-ts

Suspends TypeScript Golem agents until external events complete promises.

Official
Intermediate