ai-toolkit

Add a new rule, convention, or instruction to the project's agent configuration. Analyzes the rule and helps decide placement: root CLAUDE.md (universal rules), docs/agents/ files (topic-specific guidance), or a new skill (complex workflows). Use when users say: '/agent-add-rule', 'add a rule', 'add convention', 'new coding standard', 'add instruction for claude', 'update claude.md with'.

Add a new rule, convention, or instruction to the project's agent configuration. Analyzes the rule and helps decide placement: root CLAUDE.md (universal rules), docs/agents/ files (topic-specific guidance), or a new skill (complex workflows). Use when users say: '/agent-add-rule', 'add a rule', 'add convention', 'new coding standard', 'add instruction for claude', 'update claude.md with'.

Initialize or migrate to nested CLAUDE.md structure for progressive disclosure. Claude auto-loads CLAUDE.md from any directory it enters, so nested files get discovered automatically. Use when setting up a new project's agent config, refactoring a bloated CLAUDE.md, or adding progressive disclosure to an existing repo. Triggers on: '/agent-init-deep', 'setup progressive disclosure', 'refactor claude.md', 'split claude.md', 'claude.md is too big'.

Initialize or migrate to nested CLAUDE.md structure for progressive disclosure. Claude auto-loads CLAUDE.md from any directory it enters, so nested files get discovered automatically. Use when setting up a new project's agent config, refactoring a bloated CLAUDE.md, or adding progressive disclosure to an existing repo. Triggers on: '/agent-init-deep', 'setup progressive disclosure', 'refactor claude.md', 'split claude.md', 'claude.md is too big'.

Analyzes git diffs and commit history to intelligently fill PR templates and create pull requests via gh CLI. Use when: creating a PR, needing PR description help, saying 'create a pull request', 'fill PR template', 'make a PR', 'open a pull request'. Triggers on: create PR, open pull request, make a PR, fill template.

Analyzes git diffs and commit history to intelligently fill PR templates and create pull requests via gh CLI. Use when: creating a PR, needing PR description help, saying 'create a pull request', 'fill PR template', 'make a PR', 'open a pull request'. Triggers on: create PR, open pull request, make a PR, fill template.

Guide for creating effective, portable skills that extend Claude's capabilities with specialized knowledge, workflows, and tool integrations. Use when: (1) Creating a new skill from scratch, (2) Updating or improving an existing skill, (3) Structuring skill content for progressive disclosure, (4) Writing skill descriptions and triggers, (5) Packaging a skill for distribution. Triggers on: "create a skill", "build a skill", "new skill", "update this skill", "improve skill description", "skill structure", "skill triggers".

Guide for creating effective, portable skills that extend Claude's capabilities with specialized knowledge, workflows, and tool integrations. Use when: (1) Creating a new skill from scratch, (2) Updating or improving an existing skill, (3) Structuring skill content for progressive disclosure, (4) Writing skill descriptions and triggers, (5) Packaging a skill for distribution. Triggers on: "create a skill", "build a skill", "new skill", "update this skill", "improve skill description", "skill structure", "skill triggers".

Manage AI skills from the Ravn AI Toolkit via corvus CLI — install, update, remove, search, and configure skills for any project. Use when: (1) Installing AI skills into a project, (2) Updating installed skills to latest versions, (3) Browsing or searching available skills, (4) Configuring global or per-project skill sets, (5) Troubleshooting corvus setup. Triggers on: "install skills", "add skills", "update skills", "corvus", "skill manager", "browse skills", "set up AI rules".

Draft, evaluate, enrich, and normalize QA bug reports to RAVN standards. Trigger on "write/create a bug report", "I found a bug", "log this defect", "evaluate/score/review/improve my bug report", "normalize these bug reports", or when a user describes unexpected behavior. Also triggered by /bugreport.

Universal code quality rules — KISS, DRY, clean code, code review. Base skill every project should include. Use when writing or reviewing any code.

Universal code quality rules — KISS, DRY, clean code, code review. Base skill every project should include. Use when writing or reviewing any code.

WCAG AA and ARIA best practices — screen readers, keyboard navigation, focus management. Use when building any user-facing interface or reviewing accessibility compliance.

Visual design system patterns for web UIs. Tailwind CSS v4 design tokens and CSS variables, responsive design with container queries, dark mode, layout patterns, and spacing scales. Use when implementing visual designs, working with Tailwind CSS, or building responsive layouts.

Generate a gallery of design variations for a UI component. Takes an existing component (referenced by name, pasted code, or screenshot) and produces N distinct rendered alternatives in a single comparison page. Use when exploring visual directions, generating mockups, comparing design approaches for a component, creating A/B candidates, or when anyone says "show me options" or "give me variations" for a UI element.

Detect bundled dev work — multiple implementation tasks in one prompt — and dispatch each as an isolated agent (own session, own cwd, scoped tools, awaited), classifying dependencies so independent tasks fan out in parallel and dependent ones run in order. Use when a prompt bundles two or more dev tasks like "generate test cases and create the PR", "bug report and test plan", "do X and Y in parallel", or explicit /orchestrate. Opt into worktree mode via "--worktree" (or "in a worktree", "isolated branch", "auto-merge to main") to provision an integration branch off the protected base (develop/dev preferred; else newer main/master tip), spawn one worktree per task off that branch (parallel for independent, waves for soft-sequenced, merged-into-one for hard-chained), and hand off to agent-pr-creator for the protected-base PR. Resolves issue-tracker tickets (Linear, Jira, GitHub Issues, Notion) via MCP and uses the ticket id as the integration branch name. Skip for single-task asks, questions, or info requests.

Detect bundled dev work — multiple implementation tasks in one prompt — and dispatch each as an isolated agent (own session, own cwd, scoped tools, awaited), classifying dependencies so independent tasks fan out in parallel and dependent ones run in order. Use when a prompt bundles two or more dev tasks like "generate test cases and create the PR", "bug report and test plan", "do X and Y in parallel", or explicit /orchestrate. Opt into worktree mode via "--worktree" (or "in a worktree", "isolated branch", "auto-merge to main") to provision an integration branch off the protected base (develop/dev preferred; else newer main/master tip), spawn one worktree per task off that branch (parallel for independent, waves for soft-sequenced, merged-into-one for hard-chained), and hand off to agent-pr-creator for the protected-base PR. Resolves issue-tracker tickets (Linear, Jira, GitHub Issues, Notion) via MCP and uses the ticket id as the integration branch name. Skip for single-task asks, questions, or info requests.

Behavioral compliance testing for any CLAUDE.md or agent definition file. Auto-generates test scenarios from your rules, runs them via LLM-as-judge scoring, and reports a compliance score with per-rule pass/fail breakdown. Optionally improves failing rules via automated mutation loop. Use when: (1) testing whether your CLAUDE.md rules are actually followed, (2) evaluating an agent definition for role-boundary compliance, (3) dogfooding a skill's own SKILL.md. Triggers on: "eval", "compliance test", "test my CLAUDE.md", "check rules", "behavioral test", "/eval-agent-md". Do not trigger for: editing or writing CLAUDE.md rules, general code review, adding linting config, or any task that is not explicitly about testing behavioral compliance.

Behavioral compliance testing for any CLAUDE.md or agent definition file. Auto-generates test scenarios from your rules, runs them via LLM-as-judge scoring, and reports a compliance score with per-rule pass/fail breakdown. Optionally improves failing rules via automated mutation loop. Use when: (1) testing whether your CLAUDE.md rules are actually followed, (2) evaluating an agent definition for role-boundary compliance, (3) dogfooding a skill's own SKILL.md. Triggers on: "eval", "compliance test", "test my CLAUDE.md", "check rules", "behavioral test", "/eval-agent-md". Do not trigger for: editing or writing CLAUDE.md rules, general code review, adding linting config, or any task that is not explicitly about testing behavioral compliance.

Convert Figma component designs into production-ready React implementations with design token integration, accessibility via React Aria, and comprehensive documentation. Use when building React components from Figma designs, generating component implementation specs, or bridging design-to-development workflows.