fulcra-agent-operator

Surfaces blocked agent tasks to a human operator and relays answers back as atomic unblocks.

10|1|Updated May 21, 2026
One-click install
npx skills add https://github.com/ashfulcra/fulcra-tools --skill fulcra-agent-operator-ashfulcra
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: fulcra-agent-operator
Source: https://github.com/ashfulcra/fulcra-tools/tree/main/skills/fulcra-agent-operator
Command: npx skills add https://github.com/ashfulcra/fulcra-tools --skill fulcra-agent-operator-ashfulcra

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Agents that hit a wall often park work silently, and the human operator never finds out — the workstream is forgotten. This Skill makes operator asks first-class state on the Fulcra coordination bus, has an orchestrating agent re-surface aging asks on every heartbeat, and delivers the operator's answer as a single atomic write that unblocks the task. ## Core Features & Use Cases - Raising asks: Any agent blocks a task on the operator with coord-engine task block <team> <slug> --on-user "<ask>", writing a self-contained question with options and a default. - Orchestrator nag loop: On each heartbeat, coord-engine asks <team> --json returns waiting asks oldest-first with age_hours, so the orchestrator surfaces new asks and re-surfaces stale ones past a nag threshold. - Atomic answer leg: coord-engine answer <team> <slug> --with "<answer>" records the answer, flips the task from blocked to active, returns it to its owner's queue, and strips the needs:human tag in one write. - Use Case: An orchestrating agent polls a fulcra-agent-teams space every heartbeat, notices a task blocked for 6 hours waiting on a vault choice, pings the operator in chat, and relays the reply so the owning agent resumes on its next wake. ## Quick Start Ask the agent to check for waiting operator asks in your Fulcra team by running coord-engine asks, then surface the oldest one and relay the operator's reply with coord-engine answer.

Frequently Asked Questions about fulcra-agent-operator

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I block a task on a human operator in Fulcra?

Run coord-engine task block <team> <slug> --on-user "<ask>" with a self-contained question that includes the options, your default choice, and the consequence of waiting. Put longer context in the task body before blocking, then keep working other tasks.

How does an orchestrator agent track unanswered operator asks?

On every heartbeat, run coord-engine asks <team> --json, which returns waiting asks oldest-first with an age_hours field. Diff against previously surfaced asks, surface new ones immediately, and re-surface any ask older than your nag threshold.

What happens when the operator answers a blocked task?

coord-engine answer <team> <slug> --with "<answer>" performs one atomic write: it records the answer in next_action and the body, flips the task from blocked to active, reassigns it to its owner, and strips the needs:human tag.

Why does coord-engine answer fail on some tasks?

The answer command exits 1 if the task is not a genuine ask — meaning it is not needs:human-tagged and blocked on the operator — or if the ask has no owner. Tasks blocked on CI or another agent cannot be answered this way.

Can the orchestrator answer operator asks on its own?

No. The orchestrator is a courier, not the operator — it surfaces and re-surfaces asks and relays the human's reply, but never answers on its own authority. The answer must come from the operator through whatever channel they use.