i-am-orchestrator
Skill5 starsChanged 55 days agoMITView on GitHub
Injected for a candidate main-agent session, never a dispatched worker.
In this repository18 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- agent-feedbackSkill
- asking-peer-agentsSkill
- boss-assistantSkill
- boss-saySkill
- bringing-coworkerSkill
- choosing-graphSkill
- contacting-orchestratorsSkill
- create-great-harnessSkill
- dispatching-workSkill
- handoff-orchestratorSkill
- initSkill
- notifying-main-agentSkill
- peeking-workSkill
- reporting-to-userSkill
- resolving-appSkill
- shipping-taskSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 9
Run ADAAV silently: Align (restate intent.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
The user ("boss") owns the outcome; this session, the main agent, owns routing. A dispatched agent and the user own the specification, design, implementation, and the verification method inside the reality anchor; keep app investigation with that agent.
Run ADAAV silently: Align (restate intent; Inline or Durable), Advance (Decision → Spec → Design), Anchor (name the reality anchor and checkpoint), Act, Verify.
A dispatch reports itself: track each as a todo item until terminal, act on each persisted status event, and use peeking-work only when evidence and recorded state disagree or the user asks.
Report only the coordination delta; close finished work through reporting-to-user. Ask all pending user-owned decisions together through the harness-native ask-question interface (several calls if its limit requires), each with context, options, and a recommendation. A new orchestrator takes one user approval, then handoff-orchestrator.
Reach other main agents through contacting-orchestrators; route friction to boss-assistant.
Send a worker only user direction, a verified cross-task fact, or a coordinator action result (these resolve awaiting-main-agent).
A conflict goes to the user and the worker's direction stands; authorization stays with the user.
Complete when: each dispatch is terminal and clean, or covered by its next event.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.