Use BEFORE any `ship <project>` or `deploy <project>` command, or when the user asks to "ship" / "deploy" / "push live" a project that has a public web surface. Also use when the user asks for an "AI-SEO audit", "llm.txt", "robots.txt for ChatGPT/Claude/Perplexity", "JSON-LD", or wants to know which projects are visible to LLM crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Applebot-Extended). Generates llm.txt + sitemap.xml + robots.txt + JSON-LD schema + last_updated metadata for the project being shipped, so AI assistants cite the site in their answers.
Use BEFORE generating any non-trivial reply on Telegram (i.e., a reply that references a project, URL, person, or named entity the user cares about). Surfaces relevant project / feedback / profile / reference memory from `memory/` so the reply is grounded in current state, not assumptions. Also handles manual `/dream` invocations + scheduled 24h+5-session consolidation passes.
Use when generating a React UI component (pricing table, form, modal, dashboard tile, navbar, hero, AI-chat surface) and the goal is "novel functional component with sensible defaults". Routes to 21st.dev/magic-chat browser flow (free tier 100 credits/month, no API key needed) for the initial component, then hands the generated code back to Xantham for integration. Skip when the work is hand-styled premium taste, full-page design (use ui-ux-pro-max instead), animation-first work (use Motion or freshtechbro/claudedesignskills), or pure refactor of existing components (use frontend-design or redesign-skill).
Use BEFORE dispatching @kai on a greenfield build (no existing code) with estimated budget >4 hours. Walks GitHub spec-kit's constitution -> spec -> plan -> tasks -> analyze pipeline first, so kai gets file-based artifacts with cross-artifact consistency checks rather than a fuzzy brief. Skip for bug fixes, ops work, refactors, brownfield modifications, or any task under 4 hours of estimated work. Pinned to spec-kit v0.8.8.dev0.
Use when the user sends "sync habits", "install habits", "update habits", or any phrasing that asks for the orchestration-habits.md manifest to be applied. Fetches the latest orchestration-habits.md from the public Xantham repo, parses the embedded YAML manifest, runs every install command via Bash (skills, plugins, hooks, library files, settings.json patches), verifies each step, then reports back on the messaging channel with installed counts. Idempotent. Logs to data/xantham-habits-install.log.