triage
The triage slash command launches a diagnostic agent team to identify the root cause and impact scope of an issue without making any code changes, commits, or pull requests. Use this when you need a confidence-rated diagnosis of a bug, performance problem, or system behavior and want multiple specialized perspectives on the likely cause and blast radius before deciding whether to fix it.
mkdir -p ~/.claude/commands && curl -fsSL https://raw.githubusercontent.com/DheerG/swarms/HEAD/commands/triage.md -o ~/.claude/commands/triage.mdtriage.md
# /swarm:triage
Read `${CLAUDE_PLUGIN_ROOT}/commands/launch.md` for Step 0 (pre-flight), Step 1 (hard rules), and Step 8 (launch). This command replaces Steps 2–7.
**No lead research unless enabled.** Unless the user explicitly enables lead research, do not read codebase files, spawn subagents, or perform research. Delegate all evidence-gathering to teammates.
**Triage changes nothing.** This mode diagnoses — it does not branch, commit, edit, or open a PR. The deliverable is a confidence-rated diagnosis presented in-session.
## Settings
- **Mode:** Triage
- **Outcomes question:** "What do you want the team to diagnose? (Describe what you're seeing — the symptom, where it shows up, and any context. The team will identify the likely cause and the blast radius of fixing it, without making any change.)"
- **Defaults:** suggest-members (Triage mode), tier picked at setup (Ultra recommended), no lead research
## User-Provided Context
$ARGUMENTS
## Workflow
1. **Setup.** AskUserQuestion — "How would you like to set up the team?" Options: "Defaults — Ultra (Recommended)" (auto-configure mode/team/research; full team on the stronger model — reliable rule-following) / "Defaults — Balanced" (same auto-config; cheaper model for members — lower cost, less reliable rule-following) / "Configure each step" (choose mode, team, tier, research individually).
2. **Outcomes.** If User-Provided Context is non-empty, use as outcomes. Otherwise ask the outcomes question (plain text, not AskUserQuestion). Then echo the outcomes back verbatim — copy-paste, no condensation, no paraphrase — and use AskUserQuestion with options "Yes, that's what I meant" / "Let me add to this". If "Let me add to this", ask what they'd like to add (plain text), append it, and re-echo with AskUserQuestion until confirmed.
3. **Team configuration.** Defaults path (either Defaults option): apply the defaults above with the tier set to the option picked (Ultra or Balanced), and immediately proceed to step 4 in the same response — do not pause for user input. Configure path: follow launch.md Steps 4–6, then step 4.
4. **Confirmation.** Follow launch.md Step 7. Mode is Triage.
5. **Launch.** Follow launch.md Step 8.Read-only swarm team member. Spawned by swarm via Agent tool with team_name. Tools restricted to read and research only — no Edit, Write, or NotebookEdit. Operational guidance comes from the briefing template at spawn time.
Launch a code-mode agent team
Scaffold a custom workflow — generates a mode skill and shortcut command
Launch a general-mode agent team
Interactively launch an agent team with guided setup
Walk through swarm's core concepts and launch your first team
Run recursive refinement on the current branch and PR
Regenerate an existing custom workflow's shortcut command wiring from the current swarm template