Agent plugins, skills and hooks: adversarial code review with Codex, session handoff, and security guardrails enforced as hooks.
This repository also configures its own agents. See what agent-skills tells them →
Plugin Claude Code
Bundles 2 skills, 2 hooks · 223 tokens together
Adversarial code, architecture and specification review using OpenAI Codex as a second hostile reviewer. Two front doors (review-code, review-design) over one shared engine, plus PreToolUse hooks that enforce what a skill cannot: write-path authorisation, refutation-appendix ingestion, and rubric integrity.
Plugin Claude Code
Bundles 2 skills · 95 tokens together
Session continuity across context limits. The handoff skill writes a HANDOFF.md capturing goal, done/not-done, failed approaches and decisions with their rationale; handoff-resume reads it, drift-checks the repo, rehydrates context and continues without the user re-explaining the task.
Plugin Claude Code
Bundles 4 skills · 403 tokens together
One skill for anything a person will read. humanize classifies the text as a reply, documentation or creative copy and applies the matching ruleset (procedural-writing, descriptive-writing, creative-copy) so the reader gets it in one pass. A SessionStart hook loads the reply rules so answers in the conversation follow.