# claude-toolkit

> An opinionated Claude Code and Cursor configuration that enforces
> engineering discipline — TDD, signed commits, worktree isolation, and an
> agentic engine (kickoff enrichment, a recommendation panel, a goal
> evaluator) — through hooks and skills.

## Docs

- [README](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/README.md): documentation index
- [Getting Started](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/getting-started.md): prerequisites, installation, first workflow
- [Engineering Discipline](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/engineering-discipline.md): the six enforced ideas and their implementations
- [Architecture](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/architecture.md): coordination-branch/worktree architecture, plus the agentic engine
- [Graphify Code Graph](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/graphify.md): install the graphifyy CLI, build a repo graph, query it before grepping
- [Claude Code Configuration](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/claude-code.md): hooks, settings, plugins, policies
- [Cursor Configuration](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/cursor.md): agents, commands, rules, hooks
- [Deployment Scripts](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/deployment.md): deploy.sh, capture.sh, diff.sh
- [Skills & Commands Reference](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/skills-and-commands.md): every slash command and skill
- [Skill README Template](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/skill-readme-template.md): SKILL.md vs README.md standard
- [Testing Agent Models](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/testing-agent-models.md): verifying Cursor agent model settings
- [Agents](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/agents.md): how an agent should traverse this repo's docs
- [Orchestration: Loops → Graphs](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/orchestration.md): the prompts → loops → graphs maturity model, mapped to toolkit components

## Patterns

- [OpenRouter Free-Tier Backend](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/patterns/openrouter-free-tier.md): running the engine at $0 against OpenRouter's `:free` catalog
- [Model ID Hygiene](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/patterns/model-id-hygiene.md): single- vs double-namespace catalog IDs and the sync gate
- [Panel Config](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/patterns/panel-config.md): validate-recommendation panel config.yml walkthrough
- [Workflow Library](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/patterns/workflow-library.md): reusable named workflow DAGs shipped in .claude/workflows/ — review-verify, chief-dispatch, weekly-audit

## Examples

- [Kickoff on OpenRouter](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/examples/kickoff-openrouter.md): live kickoff enrichment run against a `:free` model, with captured output
- [Panel on OpenRouter](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/examples/panel-openrouter.md): live validate-recommendation DA panelist dispatch, with captured verdict
- [Done on OpenRouter](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/docs/examples/done-openrouter.md): live done goal-evidence evaluation, with captured JSON verdict

## Engine source

- [tool.kickoff](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/.claude/tool/kickoff.py): kickoff enrichment
- [tool.backends](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/.claude/tool/backends.py): the provider seam (build_chat_model / invoke_llm)
- [panel.dispatch](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/.claude/skills/validate-recommendation/panel/dispatch.py): validate-recommendation panel dispatch
- [done evaluator](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/.claude/skills/done/eval.py): session-goal evidence evaluation
- [panel config template](https://raw.githubusercontent.com/ArangoGutierrez/claude-toolkit/main/.claude/panel/config.yml.template): panelist config template