rust-bin

Scaffold a new domain module following the {{project-name}} spec layout. Creates all required files with correct structure and wires CLI, API, and MCP surfaces.

Specialized configuration help for Claude Code, Codex CLI, and Gemini CLI — skills, subagents, hooks, MCP servers, plugins, slash commands, permissions, settings, memory. Use whenever the user asks to add, modify, troubleshoot, or audit any agentic-CLI configuration. Trigger this skill aggressively for anything touching .claude/, .codex/, .gemini/, settings.json, config.toml, hooks.json, .mcp.json, AGENTS.md, GEMINI.md, or whenever the user mentions "Claude config", "Codex config", "Gemini config", "agent config", or any feature listed in the routing table — even if they don't explicitly ask to "configure" something.

Authoritative checklist for reviewing apps/web/ changes — fully aligned with docs/design-docs/design-system-contract.md. Covers Aurora tokens, typography ramp, elevation, radius, scrollbar, import rules, shadcn bans, TanStack Query v5, logging, and auth. Preloaded by aurora-reviewer.

Authoritative checklist for auditing parity between Claude Code, Codex CLI, and Gemini CLI configurations across skills, subagents, hooks, and MCP servers. Encodes the cross-platform mapping rules including the hook-event capability gap. Preloaded by agent-parity.

Audit the docs/references llms.txt cache for drift — stale files, orphaned entries, and (optionally) version-bumped upstream releases. Each *-llms.txt is self-describing via YAML frontmatter; this skill reads that frontmatter and cross-references it against Cargo.lock / package.json. Used by sync-stack-llms before generating new entries.

Five drift checks for detecting where the codebase has outpaced documentation — knowledge map integrity, spec.md vs xtask parity, exec-plan lifecycle drift, stale active plans, ADR coverage gaps. Preloaded by drift-detector.

Manage llms.txt reference files in docs/references/. Create new entries from GitHub repos, refresh existing ones, list current entries, or remove obsolete ones — all via the bundled script. Use this skill whenever the user mentions creating an llms.txt for a specific GitHub repo (`/manage-llms-txt jmagar/lab`, `add a docs reference for owner/repo`, `pull docs for nextjs-form-validator`), refreshing the references (`refresh the llms.txt files`, `update our docs cache`, `re-pull all the references`), or auditing what's there (`what llms.txt files do we have`, `list the references`). Also handles removal (`remove the llms.txt for foo/bar`). The script is the source of truth — invoke it; don't re-derive its logic in conversation.

Authoritative procedure for publishing an MCP server to the official MCP registry at registry.modelcontextprotocol.io — package metadata, server.json schema, authentication, validation, publish, and verification. Preloaded by mcp-publisher.

Authoritative checklist for new or modified MCP tools — token-consciousness rules, tool registration shape, and the 4-element description template. Sourced from spec.md § MCP Conventions + § Plan Integrations → MCP additions. Preloaded by mcp-tool-reviewer.

REQUIRED whenever an execution plan's real-world state diverges from its directory location. MUST run when (a) starting work that has a proposed plan, (b) finishing work that has an active plan, (c) abandoning a proposed plan, or (d) the drift-detector flags LIKELY_ACTIVE / LIKELY_DONE / STALLED. Performs `git mv` between proposed/ → active/ → completed/, updates the Status line, and fixes cross-references. Skipping this leaves the docs lying about project state.

Walk through the release-plz + cargo-dist + cocogitto release workflow. REQUIRED before tagging a release. Verifies clean state, runs all CI gates, validates the release-plz plan, walks the version bump and changelog, then drives the tag/push.

Authoritative security checklist sourced from spec.md § Security baseline (L4302-4508). Covers CSP, OAuth callback, PKCE, JWKS, /api/v1/log CSRF, /mcp auth parity, stdio MCP, redact_url, and backtrace policy. Preloaded by security-reviewer.

Run {{project-name}} xtask enforcement gates before declaring work complete. Use before finishing any Rust code change.

Mirror every CLAUDE.md in the repo to AGENTS.md and GEMINI.md siblings via symlinks, so Codex (which reads AGENTS.md) and Gemini CLI (which reads GEMINI.md) see the same agent instructions Claude does. Use this skill whenever the user mentions syncing agent-context files, mirroring CLAUDE.md, refreshing AGENTS/GEMINI symlinks, or says things like "make Codex see our CLAUDE.md", "wire up AGENTS.md", "sync the claude memory files", or after creating a new nested CLAUDE.md anywhere in the repo (e.g. crates/core/CLAUDE.md, apps/web/CLAUDE.md). Walks the entire repo, handles nested cases, and refuses to clobber real AGENTS.md or GEMINI.md files. Idempotent — safe to invoke even when all symlinks are already correct.

Wire `.agents/skills` as a symlink to `.claude/skills` so Codex and Gemini CLI can read skills from the same canonical location Claude does. Use this skill whenever the user mentions syncing skills across platforms, the `.agents/skills` directory is missing or broken, after cloning a fresh checkout, or when the user says things like "set up the skills mirror", "wire up the skills symlink", "make Codex see our skills", or "sync skills". Idempotent — safe to invoke even if the link is already in place. Also fires the underlying script automatically via PostToolUse hook on every edit under `.claude/skills/` so manual invocation is rarely needed in normal authoring flow.

Post-cargo-generate setup for a new {{project-name}} project. Walks through first-run configuration, domain replacement, and tooling bootstrap.