agent-orchestrationlisted

# Agent Orchestration Use when work spans multiple host issues or multiple coding agents and the main risk is coordination: stale branches, overlapping files, dependency ordering, or PRs/MRs becoming unmergeable as the default branch moves. This skill coordinates work. It does not replace `task-driven-development`. Resolve `engineeringHost` via `.skills/profile` or [`config/tracker-profiles/`](../../config/tracker-profiles/README.md). Read the target repo `AGENTS.md` for worktree helpers and hub-file rules. ## Goals - Maximise safe parallelism. - Keep every agent isolated in its own git worktree and branch. - Keep feature branches current with `origin/<default>`. - Do not build on another agent's unmerged branch unless the issue declares that dependency and it is closed/merged. - Keep CI and mergeability authoritative before handoff. - Remove stale worktrees after merge or abandon. ## Repository workflow 1. Never work directly on `main` / `master`. 2. Fetch origin before starting. 3. Create each agent branch/worktree from current `origin/<default>`. 4. Before push: refresh from default branch, resolve conflicts per `AGENTS.md`, run relevant tests. 5. Open/update PR (GitHub) or MR (GitLab) with closing keyword linking the issue. 6. Wait for required CI; address feedback on your own PR/MR only. ## Branch and worktree naming Prefer: ```text <type>/<issue-number>-<kebab-title> ``` Use the repo's worktree helper when present (for example `pnpm worktree:add`); otherwise `