orch

A command that coordinates several AI helpers on a larger task. It can delegate work, monitor progress, choose an execution mode, and compress the conversation when it becomes too long.

In plain words
What is it for?
Use it for multi-step development tasks, preview an orchestration plan, run helpers in separate workspaces, view status or a timeline, continue a session, or abort it.
Why use it?
It helps divide complex work into smaller assignments and track what each helper is doing. Preview and control commands let you inspect, continue, or stop the work.

Command

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/data-wise/craft/orch
Clone the repo
git clone --depth 1 https://github.com/Data-Wise/craft
Per session 18 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,467 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00018 $0.03467
Opus 5 $0.00009 $0.01733
Sonnet 5 $0.00004 $0.00693
Haiku 4.5 $0.00002 $0.00347

Measured yesterday against content hash b75cf2ccf995, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

orch scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

commands/orch.md · 371 lines

How it starts

The opening of the file, as written. The whole thing — 371 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/craft:orch — Launch Orchestrator Mode

Usage

/craft:orch <task>              # Start with default mode
/craft:orch <task> <mode>       # Start with specific mode
/craft:orch <task> --dry-run    # Preview orchestration plan
/craft:orch <task> -n           # Preview orchestration plan
/craft:orch <task> --swarm     # Isolated worktrees per agent
/craft:orch status              # Show agent dashboard
/craft:orch timeline            # Show execution timeline
/craft:orch compress            # Force chat compression
/craft:orch continue            # Resume previous session
/craft:orch abort               # Stop all agents

--refine (prompt pre-processing)

When --refine is set, do NOT act on the raw argument. First invoke the prompt-refiner skill with the argument and project context. Follow that skill's canonical flow (before/after box → fenced refined-prompt block → 4-way confirm: Execute now / Copy for elsewhere / Edit first / Skip; --yes or auto mode auto-accepts Execute now). If the user picks "Copy for elsewhere," stop here — do not start orchestration. Otherwise proceed using the prompt the skill returns. On no-argument interactive commands, refine AFTER the topic is captured.

--yes cascades: the prompt-refiner auto-accepts AND the interactive loop is suppressed — one flag, fully headless.

Dry-Run Mode

Preview the orchestration plan without spawning any agents:

┌───────────────────────────────────────────────────────────────┐
│ 🔍 DRY RUN: Orchestrator v2.1                                 │
├───────────────────────────────────────────────────────────────┤
│                                                               │
│ ✓ Task Analysis:                                              │
│   - Input: "add user authentication with OAuth"               │
│   - Complexity: Complex                                       │
│   - Mode: default (2 agents max)                              │
│   - Estimated subtasks: 5                                     │
│   - Delegation strategy: Hybrid (parallel + sequential)       │
│                                                               │
│ ✓ Orchestration Plan:                                         │
│                                                               │
│   Wave 1 (Parallel - 2 agents):                               │
│   ├─ Agent: arch-1 (architecture)                             │
│   │  Task: Design OAuth flow and security model               │
│   │  Estimated: ~8 minutes                                    │
│   │  Dependencies: None                                       │
│   │                                                           │
│   └─ Agent: doc-1 (documentation)                             │
│      Task: Research OAuth 2.0 best practices                  │
│      Estimated: ~5 minutes                                    │
│      Dependencies: None                                       │
│                                                               │
│   Wave 2 (Sequential - awaits Wave 1):                        │
│   ├─ Agent: code-1 (backend)                                  │
│   │  Task: Implement auth endpoints                           │
│   │  Estimated: ~15 minutes                                   │
│   │  Dependencies: arch-1                                     │
│   │                                                           │
│   ├─ Agent: code-2 (frontend)                                 │
│   │  Task: Create login/logout UI                            │
│   │  Estimated: ~12 minutes                                   │
│   │  Dependencies: arch-1                                     │
│   │                                                           │
│   └─ Agent: test-1 (testing)                                  │
│      Task: Generate test suite                                │
│      Estimated: ~10 minutes                                   │
│      Dependencies: code-1, code-2                             │
│                                                               │
│ ✓ Resource Allocation:                                        │
│   - Max concurrent agents: 2                                  │
│   - Total agents required: 5                                  │
│   - Estimated total time: ~35 minutes (with parallelization)  │
│   - Sequential time: ~50 minutes                              │
│   - Time saved: ~15 minutes (30%)                             │
│                                                               │
│ ⚠ Warnings:                                                   │
│   • Context usage will be monitored (compression at 70%)      │
│   • Progress dashboard updates every 30 seconds               │
│   • Session state auto-saved at checkpoints                   │
│                                                               │
│ 📊 Summary: 5 agents, 2 waves, ~35 min execution              │
│                                                               │
├───────────────────────────────────────────────────────────────┤
│ Run without --dry-run to execute                              │
└───────────────────────────────────────────────────────────────┘

Read the full file on GitHub · 371 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 371 lines · 18 tokens per session scan A b75cf2ccf995

Subscribe to this mod's changes

orch is a command published in the GitHub repository Data-Wise/craft (4 stars, last pushed 16d ago), licensed MIT. It adds 18 tokens to every session and 3,467 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.