← ClaudeAtlas

multi-project-contextlisted

Resolve a task to exactly one project root and keep each project's context isolated, covering the one-project-per-worker model. Use before delegating whenever a request names a project, spans or touches multiple repositories, uses an informal or ambiguous project name, or could route a worker's context into the wrong project.
mohammadhprp/teammate · ★ 1 · AI & Automation · score 74
Install: claude install-skill mohammadhprp/teammate
# Multi-project context One primary agent coordinates several independent projects, and context must not leak between them. Resolve the project before delegating so every worker sees exactly one project's context. ## When to use - A request names a project, a repository, or several projects. - A task spans repositories, or workers from different projects run at once. - A project name is informal or ambiguous and must be mapped to a root. ## When not to use - The project is already resolved for this task and its worker is running in it. ## Resolving a project - Map each project to its root directory and its registered name. The name defaults to the `--root` basename; pass `--name` to override it. - Register the resolved project so its identity is durable, not re-typed: `python3 scripts/tm.py project add --name <name> --root <root>`. Use `python3 scripts/tm.py project list` to reuse a known root, and to catch a name that already points somewhere else — the registry refuses to rebind a name to a different root without `--force`. - Before delegating, prepare it: `bootstrap-project` writes its `AGENTS.md`, syncs the harness's skills and agent definitions, and installs the skills the work needs, so the worker starts with context. - When the developer uses an informal name, or two projects look similar, confirm the resolved root before delegating. Do not guess. - A task may touch several projects, but each worker still has one project scope. State explicitly