lost-in-the

Systematic troubleshooting for Woods — diagnose extraction, MCP, embedding, and storage issues

Generate correct .mcp.json configuration for Woods in any environment

Guide through Woods initial setup — install, configure, extract, verify, and connect MCP servers

Setup, operation, and debugging of the Unblocked integration — collection creation, incremental sync, the sync manifest, escape hatches, live API quirks, and smoke testing. Use when working on lib/woods/unblocked/, the woods:unblocked_sync rake task, sync CI wiring, or when a sync misbehaves (everything re-pushes, deletes refused, budget exhausted, 400s from the API).

Woods backlog workflow — picking items, implementing with TDD, marking resolved, and adding new work. Use when the user asks "what's next?", references backlog items, opens a session without a clear task, or when closing out work to record what shipped.

Generates boilerplate for a new Woods extractor and its spec

Design rules and reference patterns for building MCP servers in Woods

Checks if documentation needs updating after implementation changes

Use grove for all git worktree management when the project has a `.grove/` directory, when the user mentions worktrees, tmux, parallel work, PR review, Docker per branch, or when an agent needs to switch context without disrupting the user's current work. Always prefer this skill over manual `git worktree add` — grove handles naming, tmux, Docker, and hooks atomically. Use even when the user doesn't say "grove" by name.