codeArbiter
Enforced test-first gates, reviews, and audit trails for coding
All Skills in This Repository (23)
Pure Emerald Level Indicatorsusing-git-worktrees
Create isolated git worktrees and branches for parallel agent tasks.
skill-author
Author v2-format codeArbiter skills after validating gaps against the existing index.
decision-variance
Compare architectural artifacts against the scaffold and label variances as concur, divergent, or silent.
dispatching-parallel-agents
Coordinate parallel agent fan-out and collect terminal results without dropping failures.
executing-plans
Execute approved implementation plans in small user-checked batches.
release
Derive SemVer from Conventional Commits and publish annotated tags and GitHub Releases.
finishing-a-development-branch
Resolve development branches through PR-only terminal actions with gate verification.
security-architecture
Map attack surfaces and identify unmitigated STRIDE threats before implementation.
secret-handling
Detect and block secret leaks in code changes.
brainstorming
Refine rough feature ideas into approved, testable specifications with acceptance criteria.
decision-lifecycle
Author and track numbered, dated architecture decision records with user attribution.
commit-gate
Enforce a nine-phase commit workflow with tests, linting, and secrets scanning.
Frequently Asked Questions
FAQPage SchemaHow to install codeArbiter?βΌ
Run `npx skills add arbiterForge/codeArbiter --all -g -y` in your terminal to install all skills globally. Then open a repository and run the init command to opt it into governance.
What does codeArbiter actually enforce?βΌ
It blocks feature code until a spec is approved and a failing test exists, blocks commits on failing tests or leaked secrets, and requires review findings to clear before a PR opens. Every decision and override is logged for audit.
Does codeArbiter work with Claude Code and Codex?βΌ
Yes. It ships sibling plugins for Claude Code, Codex CLI, and Pi, all generated from one shared core, so a repository keeps its specs, plans, and decisions when switching hosts.
Can codeArbiter run autonomously?βΌ
Yes. The sprint mode executes an approved plan end to end without per-batch checkpoints, but hard gates like security findings and merges always stop for the user.
Do I need to change how my project is structured?βΌ
No. It adds a single .codearbiter directory to your repository for state and stays dormant until you explicitly opt the project in.
Related Repositories in Software Engineering
View All in Software Engineeringβopenclaw
Run a personal AI assistant across your devices and chat apps
superpowers
Gives coding agents a disciplined workflow from idea to merged code
react
AI agent skills for building, testing, and porting React core