-v, --verbose | Verbose output |
-d, --dry-run | Show what would be run without executing |
-a, --agent <agent> | Run specific agent within squad |
-t, --timeout <minutes> | Per-agent execution timeout in minutes (default: 15) |
-p, --parallel | Run all agents in parallel (add -b to run them in the background) |
-l, --lead | Lead mode: single orchestrator using Task tool for parallelization |
-b, --background | Run agent in background (detached process) |
-w, --watch | Run in background but tail the log for visibility |
--use-api | Use API credits instead of subscription |
--effort <level> | Effort level: high, medium, low (default: from SQUAD.md or high) |
--skills <skills...> | Skills to load (skill IDs or local paths) |
--provider <provider> | LLM provider: anthropic, google, openai, mistral, xai, aider, ollama |
--model <model> | Model to use (e.g., opus, sonnet, haiku, gemini-2.5-flash, gpt-4o) |
--trigger <type> | Trigger source: manual, scheduled, event, smart (default: manual) |
--cloud | Dispatch execution to cloud worker via API (requires squads login) |
--task <directive> | Founder directive for conversation mode (replaces lead briefing) |
--max-turns <n> | Max conversation turns (default: 20) (default: 20) |
--cost-ceiling <usd> | Cost ceiling in USD (default: 25) (default: 25) |
--no-verify | Skip post-execution verification (Ralph loop) |
--execute | Explicitly execute agents (default for run <target>) |
-j, --json | Output as JSON |
-i, --interval <minutes> | Autopilot: minutes between cycles (default: 30) |
--max-parallel <count> | Autopilot: max parallel squad loops (default: 2) |
--budget <usd> | Autopilot: daily budget cap ($) (default: 0) |
--once | Autopilot: run one cycle then exit |
--phased | Autopilot: use dependency-based phase ordering (from SQUAD.md depends_on) |
--no-eval | Skip post-run COO evaluation |
--org | Run all squads as a coordinated org cycle (scan → plan → execute → report) |
--force | Force re-run squads that already completed today; bypasses pause enforcement |
--resume | Resume org cycle from where quota stopped it |
--wait-for-quota | Org cycle: on quota cap, poll until the session window reopens instead of stopping |
-y, --yes | Skip the org-run cost confirmation (for deliberate/non-interactive triggers) |
--focus <mode> | Cycle focus: create, resolve, review, ship, research, cost (default: create) |
