Claude Code & OpenClaw plugin, with a context-aware rules engine that injects only the rules each prompt needs. Codebase-aware, ~2ms lookup, pure Python. 222 rules / 30 domains for full-stack and scientific work. Includes a deterministic security audit, automatic session handoffs, ranked project memory, context usage alerts, and review agents.
Agent
Devil's advocate for architecture decisions. Use when proposing a new system design, choosing a technology, or making a significant structural change. Challenges assumptions and proposes alternatives.
Agent
Adversarial code reviewer. Use when you want a brutally honest review of code quality before a PR, merge, or deployment. Focuses on bugs, performance, maintainability, and missed edge cases. Read-only.
Agent
Analyzes code for performance issues. Focuses on N+1 queries, unnecessary re-renders, memory leaks, bundle size, and slow algorithms. Use before deployment or when something feels slow.
Agent
Analyzes code for refactoring opportunities. Identifies code smells, duplications, overly complex functions, tight coupling, and suggests specific improvements. Read-only — reports findings, doesn't modify.
Agent
Use after the red team has reported findings. Proposes concrete fixes, hardening measures, and defense-in-depth strategies. Reviews red team findings for false positives and prioritizes remediation by risk.
Agent
Use when reviewing code for security vulnerabilities. Thinks like an attacker. Searches for current CVEs and OWASP Top 10 issues relevant to the tech stack. Invoke for security audits, pre-deployment reviews, or when touching auth, payments, or user data.
Agent
Generates comprehensive tests for code. Reads existing test patterns in the project and matches the style (Jest, Vitest, pytest, etc.). Covers happy paths, edge cases, error cases, and boundary conditions.