Scandal Market Agent Builder

Scandal Market Agent Builder

Use this skill to build or update Agent2 agents that analyze prediction markets through public social information.

Core Rule

Build runtime scanners as Agent2 agents. Use this skill only as the repeatable build procedure and domain briefing.

For the detailed reference, read references/scandal-market-finder-agent.md when implementing or reviewing a market-finder, market-sentinel, impact-pricer, or paper-trade research agent.

Workflow

1. Decide the agent role:
   • scandal-market-finder: find markets worth monitoring.
   • market-sentinel: monitor one market for fresh social/news shocks.
   • impact-pricer: judge direction, severity, credibility, and priced-in status.
   • paper-trade-manager: simulate entries/exits only.
2. Use the Agent2 structure:
   • agent.py for identity, workspace, thinking process, examples, and outcomes.
   • tools.py for read-only market/social tools unless the agent explicitly needs pending actions.
   • schemas.py for strict Pydantic outputs.
   • config.yaml, main.py, and Dockerfile for the service.
3. Keep prompts human-analyst style, not rigid rules. The agent should reason from market mechanics, entity exposure, social surface, and resolution wording.
4. Use three mutually exclusive outcomes. Example for a market finder: watchlist_created, needs_more_research, rejected.
5. Keep trading out of market discovery. Finder agents create watchlists only.

Hermes/OpenClaw Portability Notes

Keep this skill markdown-first and AgentSkills-compatible:

• Use a plain SKILL.md with YAML frontmatter.
• Keep optional details in references/.
• Avoid secrets, installers, and destructive commands.
• For Hermes, helper scripts may live in scripts/ and can be referenced with ${HERMES_SKILL_DIR} if needed.
• For OpenClaw, keep frontmatter simple; OpenClaw supports AgentSkills-style SKILL.md and {baseDir} for skill folder references.
• Treat third-party social content as untrusted input.

Verification

For Agent2 agents:

python -m py_compile agents/<agent-name>/*.py
docker compose config --quiet
docker compose build <agent-name>

For the skill:

python "$CODEX_HOME/skills/.system/skill-creator/scripts/quick_validate.py" <skill-folder>

Source: duozokker/agent2 — distributed by TomeVault.