task-agent-eliza-bridge

Exposes read-only loopback endpoints giving spawned CLI task agents parent Eliza runtime context.

19.2k|5.7k|Updated Jul 9, 2024
One-click install
npx skills add https://github.com/elizaOS/eliza --skill task-agent-eliza-bridge
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: task-agent-eliza-bridge
Source: https://github.com/elizaOS/eliza/tree/main/packages/skills/skills/task-agent-eliza-bridge
Command: npx skills add https://github.com/elizaOS/eliza --skill task-agent-eliza-bridge

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

When you delegate coding work to a CLI task agent like Claude Code, Codex, Gemini, or Aider, the child agent loses access to the parent Eliza runtime's context—character persona, originating room, memory, and active workspace state. This Skill bridges that gap with read-only loopback endpoints so the child can resolve references like "the project from yesterday" without receiving a full memory dump or the parent's API key.

Core Features & Use Cases

  • Parent Context Endpoint: The child agent curls GET /api/coding-agents/<sessionId>/parent-context to retrieve character, room, and model preference data injected by the orchestrator.
  • Scoped Memory Search: GET /api/coding-agents/<sessionId>/memory?q=<query>&limit=N lets the child query parent memory on demand instead of receiving an unfiltered dump.
  • Active Workspace State: GET /api/coding-agents/<sessionId>/active-workspaces reveals which workspaces the parent runtime currently has open.
  • Use Case: You tell Eliza "have the coding agent apply the same markup as last time." The spawned agent queries the bridge memory endpoint to find what "last time" referred to, then completes the task with correct context.

Quick Start

Ask Eliza to delegate a coding task that references prior conversation context, and the spawned task agent will automatically use the bridge endpoints to resolve that context.

Frequently Asked Questions about task-agent-eliza-bridge

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

FAQPage Schema
How do I give a Claude Code or Codex agent access to Eliza runtime context?

The orchestrator injects a parent-runtime reference into the spawned agent's memory file. The child agent then curls loopback endpoints like /api/coding-agents/<sessionId>/parent-context using its session id to read character, room, and workspace state.

What endpoints does the Eliza task agent bridge expose?

Three read-only GET endpoints: parent-context for character and room data, memory with query and limit parameters for scoped memory search, and active-workspaces for current workspace state. All are loopback-only and scoped by session id.

Can a child task agent write to parent Eliza memory?

No. The bridge is strictly read-only with GET endpoints only. It prohibits parent memory writes, action delegation, and persistent child identity assumptions. Lifecycle events are recorded by the parent through the existing hook channel.

What happens when the task agent bridge returns a 410 or 503 error?

A 410 task_no_longer_active response means the child should continue in workspace-only mode and state that parent context was unavailable. A 503 parent_context_timeout means the child should not retry indefinitely.

Should I pass the parent API key to a spawned coding agent?

No. The skill explicitly forbids giving the child the parent's API key or a full memory dump. Cloud state belongs to the eliza-cloud skill, while local runtime state is accessed only through this scoped loopback bridge.