A portable configuration and tooling kit for AI coding agents - Claude Code, OpenAI Codex, and OpenCode.
This repository also configures its own agents. See what agents tells them →
Skill Claude CodeCodex
Write idiomatic, production-grade Go. Self-contained — all rules inline, no reference files. Use when editing .go files, go.mod, go.sum, or Go tests, or when asked to design, generate, or review Go code. Covers style, control flow, errors, concurrency, context, testing, performance, security, modules, JSON, database…
Skill Claude CodeCodex
Review the current branch's changes against its base with multiple AI agents. Each finding carries a paste-ready PR comment; pass no-pr to omit it.
Skill Claude CodeCodex
Produce a structured implementation plan document and save it to disk. Use when asked to write, create, or generate an implementation plan for a feature or change.
Skill Claude CodeCodex
Implement a plan written by the plan-doc skill (.agents/plans/.md) on its feature branch, one commit per logical change. Use only when explicitly asked to implement, execute, or carry out such a plan.
Skill Claude CodeCodex
Review a plan doc written by the plan-doc skill with multiple AI agents, verify their findings, and apply the verified fixes to the plan file in place. Use when asked to review, audit, or critique a plan written by plan-doc.
Skill Claude CodeCodex
Stepped git flow — create a feature branch, commit staged changes, then push and open a PR. Run all steps or select a subset by number or name (branch, commit, pr).
Skill Claude CodeCodex
Default-branch resolution, branch naming and creation, Conventional Commits, commit hygiene, and push, pull request, and attribution rules. Use before creating a branch, committing, pushing, or opening a pull request.
Skill Claude CodeCodex
Shared protocol for fanning a review out to the ask-agent agents — prompt file, agent selection, background launch, retry by resume, response classes, the orchestrator's own review, verification and merging, coverage, and stopping calls. Load only when a review skill (multi-code-review, plan-review) says to.
Skill Claude CodeCodex
Ask another agent's CLI (claude, codex, glm, kimi, …) for a read-only second opinion, review, or research, and continue that conversation later. Use when the user asks to consult another agent or model, or when another skill says to. Never use it when you were invoked through ask-agent yourself.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: