Autonomous Agents
Building applications
Use autonomous agents plus skills to create end-to-end Lucid applications.
This guide shows a compact workflow for using an AI agent to build full Lucid applications.
1. Install the Lucid skill
mkdir -p .claude/skills/lucid-agent-creator && \
curl -fsSL https://raw.githubusercontent.com/daydreamsai/skills-market/main/plugins/lucid-agent-creator/skills/SKILL.md \
-o .claude/skills/lucid-agent-creator/SKILL.md2. Prompt for the full app, not a single function
Use prompts that request:
- Agent goal and user workflow
- Entrypoints with input/output schemas
- Pricing (if paid)
- External API constraints (
allowedHosts)
Example prompt:
Build a Lucid application called "research-assistant".
Create an agent with:
- summarize(text: string) -> { summary: string }
- extract_entities(text: string) -> { entities: string[] }
Make summarize paid at 0.01 USDC.
Use valid JavaScript handler code and call create_lucid_agent.
Return the invoke URL and sample curl commands.3. Iterate by prompting changes
Ask the same agent to make incremental updates:
- Add or remove entrypoints
- Tune pricing and payment policy
- Add identity metadata and registration settings
4. Validate behavior
After generation, invoke your created endpoints and verify:
- Schemas validate expected input/output
- Paid endpoints enforce x402 payment requirements
- Hosted URLs and metadata are returned correctly