Starts creating a new Claude Code environment for any project -- interviews the user, generates CLAUDE.md + rules + agents + skills + settings, and validates the result. Supports both single-project environments and multi-area hubs (several related work areas sharing a parent configuration layer). Auto-detects when run inside an existing hub and offers to add a new area. Use when user says "create environment", "set up my project", "generate environment", "I need a Claude setup", "make me an environment", "add a work area", "/create", or "build me an environment". Do NOT use for updating existing environments, validating existing environments, or best-practice upgrade recommendations.
Updates the Harness Generator's own knowledge base (Docs/AgentGuidelines/Topics/, AgentPlaybooks/, tool-registry) -- ingests pending ProvideKnowledge/ docs, then web-researches Anthropic docs and changelogs, incorporates validated findings with source attribution, and prunes superseded or deprecated entries (90-day tool-registry cadence). Supports a local-only mode that ingests ProvideKnowledge/ without web search. Use when the user says "update knowledge", "refresh best practices", "check for updates", "update the Harness Generator", "process knowledge", "I added some docs to ProvideKnowledge", "incorporate this research", "process my notes", or "/update". Do NOT use to upgrade a user's generated environment (use /upgrade-environment) or to validate one (use /validate-environment).
Audits an existing Claude Code environment against current best practices, interviews the user about pain points, and implements approved improvements -- including shape conversions (single <-> multi-area hub <-> undeclared siblings). Recommendations are tiered quick wins / medium / large so the user picks what to implement. Use when user says "upgrade environment", "improve my environment", "optimize my setup", "audit my environment for best practices", "make my environment better", "convert to hub", "collapse hub", "/upgrade-environment", or "what can I improve". Do NOT use for creating new environments or for structural-only validation (use /validate-environment for the latter).
Validates an existing Claude Code environment against structural correctness and quality standards (file references resolve, frontmatter is valid, size limits hold, routing is complete, hub registry matches disk). Use when user says "validate environment", "check my environment", "is my setup correct", "validate my config", "audit my environment", "run the validator", or "/validate-environment". Do NOT use for creating new environments or for best-practice upgrade recommendations (use /upgrade-environment for the latter).