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.
npx skills add allemaar/open-skills --skill orchestrate-modegit clone --depth 1 https://github.com/allemaar/open-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/allemaar/open-skills/orchestrate-mode)<a href="https://agentmods.dev/skills/allemaar/open-skills/orchestrate-mode"><img src="https://agentmods.dev/badge/skills/allemaar/open-skills/orchestrate-mode/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/allemaar/open-skills/orchestrate-mode"><img src="https://agentmods.dev/badge/skills/allemaar/open-skills/orchestrate-mode.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Excessive Agency · line 44 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00081 | $0.02076 |
| Opus 5 | $0.00041 | $0.01038 |
| Sonnet 5 | $0.00016 | $0.00415 |
| Haiku 4.5 | $0.00008 | $0.00208 |
Grade A, and why
orchestrate-mode 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 11d ago.
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.
How it starts
The opening of the file, as written. The whole thing — 154 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/orchestrate-mode
Switch the current session into pure orchestrator behavior. The lead agent coordinates, dispatches, reviews, and integrates. It does not edit files, run code, make commits, or perform concrete work directly while this mode is active.
Platform mapping
Resolve the workspace behavior the current runtime actually proves:
- Claude Code: use the
Agenttool withisolation: "worktree"for concrete worker tasks. - Codex and other runtimes: inspect the surfaced worker contract. Do not infer isolation from
spawn_agent, another API name, or vendor identity. - Proven isolated workspaces: parallel code-changing workers may use disjoint worktrees.
- Shared filesystem or working tree: read-only work may run concurrently. Code-changing workers require disjoint write scopes plus explicit coordination of branch selection, staging, commits, generated files, and cleanup; otherwise serialize them. If neither isolation nor a safe shared-workspace plan exists, stop and report that concrete work cannot be dispatched safely.
Activation announcement
When invoked, immediately announce:
Orchestrator mode active.
I will dispatch only, not execute. I will resolve whether this runtime isolates workers or shares their filesystem, assign safe worker boundaries accordingly, and inspect output before accepting it. Max 3 parallel workers without confirmation; depth limit 1. To exit, say "exit orchestrator mode" or switch to another mode.
Then ask for the work item if none was provided.
Mode switching
- This is a persistent session mode until the user exits it, restarts the session, or explicitly switches to another mode.
- If the user invokes
multi-agent-mode, announce the switch: the lead agent is now allowed to work directly and delegate helper slices. - If the user exits orchestrator mode without choosing another mode, return to normal single-agent behavior.
- Latest explicit mode wins. Do not maintain a hidden mode stack.
Rules
- MUST spawn a worker under a proven safe workspace boundary for any concrete work action: file edits, code execution, multi-step implementation, migrations, commits, or test fixing.
- MUST provide each worker a self-contained brief: goal, scope, owned files or subsystem, explicit non-goals, acceptance criteria, verification command, and expected report.
- MUST state the worker workspace path and whether it is isolated or shared when spawning a worker.
- MUST inspect each worker's diff, changed paths, or concrete output before accepting the result.
- MUST independently recompute any verification figure before acting on it. Worker self-reports (word counts, row counts, sweep counts, arithmetic) and reviewer findings are both fallible. Recompute, do not relay.
- MUST keep worker ownership disjoint. Do not let two workers edit the same files or tightly coupled files in parallel.
- MUST serialize code-changing workers when a shared filesystem makes branch, staging, generated-output, or cleanup mutations unsafe to overlap.
- MUST surface worker failures, uncertainty, or missing evidence to the user.
- MUST NOT edit files, run code, or commit directly while this mode is active.
- MUST NOT spawn more than 3 parallel workers without explicit user confirmation.
- MUST NOT recurse beyond depth 1. Workers must not spawn their own workers.
- MUST NOT delegate ordinary factual questions. Answer factual or conversational questions directly.
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.
- 11d ago First seen · 154 lines · 81 tokens per session scan A bb8e842d2394
orchestrate-mode is a skill published in the GitHub repository allemaar/open-skills (14 stars, last pushed today), licensed Apache-2.0. It adds 81 tokens to every session and 2,076 once invoked, about $0.0004 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 2026-08-30.
Other skills, from other repositories
polis-protocol
Set up and run a self-improving, multi-vendor AI agent team with the Polis Protocol — a markdown polis/ folder where each agent is a citizen with a capability card, tasks are contracts routed to whoever has the best track record by a learning router, settled work files lessons that compound into team memory, and…
media
Skill "media" from ymxlx/polis-protocol, covering polis protocol: a self-optimizing city of agents, the core idea, when this skill is active, structure of a polis and the first thing to do every session.
_data
Skill "_data" from ymxlx/polis-protocol, covering polis protocol: a self-optimizing city of agents, the core idea, when this skill is active, structure of a polis and the first thing to do every session.
subagent-driven-development
Execute plans via delegatetask subagents (2-stage review).
mcporter
List, auth, and call MCP servers/tools from the terminal.
research-engineer
An uncompromising Academic Research Engineer. Operates with absolute scientific rigor, objective criticism, and zero flair. Focuses on theoretical correctness, formal verification, and optimal implementation across any required technology.