setup-agent-teamlisted

# Setup Agent Team ## Overview Use this skill to decide whether agent-team delegation is warranted and, when it is, produce a tool-neutral manual execution packet. Do not launch subagents, run runtime-specific workers, or edit the target repo as part of setup. ## Trigger Gate Use an agent team only when all are true: - The work spans multiple domains, surfaces, or viewpoints. - Meaningful parts can proceed in parallel. - A single agent's context would become too large, tangled, or risk-prone. Refuse or downgrade for: - Small bugs - Single-file edits - Clear one-slice features - Narrow docs/spec/glossary edits - Tasks with no meaningful parallel ownership - Requests that need broad write access before scope is known - Private-vault, credential, production, or package-fetching work without explicit approval If refusing, output a short downgrade decision and recommend the appropriate single-agent workflow. ## Workflow 1. Inspect the user goal and available repo/project context. 2. Record the project baseline: path, repo guide, worktree state, known user changes, existing verification, and whether this continues an existing project or starts a new one. 3. Apply the trigger gate. Refuse if the task does not meet it. 4. Select only roles that own distinct questions, artifacts, or verification responsibilities. 5. Define context packets for each selected role. 6. Define ownership boundaries, read/write scopes, and disallowed paths. 7. Identify external systems a