OpenParable
Skip to content
← Back to skills

Eliza

Agent Building
eliza

Eliza is an Agent Skill for agent building published in the elizaOS/eliza repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Use for spawned coding/task agents working on the Eliza app or asking what the running Eliza agent can do. Covers the repo CWD, default skill loading and overrides, child-to-parent USE_SKILL calls, parent runtime context APIs, plugin loading/building, Cloud app/payment/domain/media capabilities, and how workers can ask the parent agent to use loaded capabilities.

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/elizaOS/eliza/blob/HEAD/packages/skills/skills/eliza/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/eliza, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

Eliza

Eliza is this repo's local-first agent app and Cloud-backed product built on elizaOS. Use this skill when a worker needs to understand the running Eliza agent, work inside the Eliza checkout, build or modify the agent, or ask the parent runtime for information/actions that are not available from files alone.

Read These References First

  • references/agent-orchestration.md for spawned-worker protocol, CWD, SKILLS.md, USE_SKILL, and parent runtime APIs
  • references/capability-map.md for what Eliza can do through plugins, services, actions, connectors, Cloud, and orchestration
  • references/plugin-and-skill-lifecycle.md for loading, unloading, building, and overriding plugins and skills

Worker Protocol

  1. Read the local repo instructions first: AGENTS.md, CLAUDE.md, and any task-provided memory file.
  2. Stay in the provided CWD. The orchestrator injects the workspace path into the agent memory file; do not leave it for /tmp, $HOME, or another checkout.
  3. If ELIZA_SKILLS_MANIFEST or SKILLS.md exists, read it before assuming a capability is unavailable.
  4. To ask the running parent Eliza agent for information or actions, emit a standalone line:
USE_SKILL parent-agent {"request":"Describe exactly what you need the parent Eliza agent to do"}
  1. To inspect parent actions first:
USE_SKILL parent-agent {"mode":"list-actions","query":"github"}
  1. To call deterministic Eliza Cloud APIs through the parent account, inspect commands with USE_SKILL parent-agent {"mode":"list-cloud-commands"} and call them with mode:"cloud-command". Paid, mutating, or destructive Cloud commands require parent/user confirmation and confirmed:true.
  2. Use parent capabilities for data/actions that belong to the parent runtime: calendar, GitHub, browser, app/cloud account state, user-confirmed paid operations, payment requests, domain purchases, promotion/media generation, private memory, connector state, and plugin-specific tools.
  3. If the parent asks for confirmation or returns a blocker, continue from that result. If you need the human to decide, print DECISION: <specific question or blocker>.

Related Skills

  • elizaos for upstream runtime abstractions and plugin development
  • eliza-app-development for this app repo's layout and product architecture
  • eliza-cloud for Cloud APIs, app auth, containers, credits, billing, payments, domains, promotion, and media generation
  • build-monetized-app for shipping Cloud apps that earn from inference usage, app-credit purchase share, charge links, or x402 payment requests
  • task-agent-eliza-bridge for lower-level child-to-parent bridge details