← ClaudeAtlas

activationlisted

Create and verify the harness-appropriate per-project activation file that arms atelier's hooks. Use when someone asks to turn on, configure, or check atelier enforcement (custody, worker context, worktree isolation, handoff routing) in a project, or when a hook that should be firing appears silent.
hsb3/dotfiles-agents · ★ 0 · AI & Automation · score 70
Install: claude install-skill hsb3/dotfiles-agents
# Activation Atelier keeps `atelier.local.md` in the sole configured agent's native directory (`.claude`, `.codex` or `.opencode`), or `.agents` when multiple agents are configured. Native directories and root `opencode.json`/`opencode.jsonc` files identify agents; installed binaries, instruction files and `.agents` itself do not. Symlinks are not configuration markers. With no configured agent, creation defaults to the current harness. `ATELIER_ACTIVATION_FILE` is authoritative, relative to the project root when relative. Otherwise readers select an existing `.agents` policy first, the canonical location next, then legacy `.claude`, `.codex`, `.opencode` locations in that fixed order. A selected malformed file never falls back. A worktree's local policy wins before main-checkout inheritance. Runtime readers never migrate files. `create` and setup move an existing policy without changing its bytes or permissions, coalesce byte-identical duplicates, and reject divergent policies before any mutation, even with `--force`. Repeating `create` leaves an existing canonical policy unchanged. Existing handoff paths and stamps remain unchanged. It is the one file that arms atelier's enforcement layer. Absent, it means everything is off. This skill creates it and tells you whether it is actually doing anything — the two states look identical from the outside otherwise, because every loader in atelier fails open (`except Exception: return <inert default>`) by design. ## Procedure Ru