The Agent Skills Directory

Structured Data Discovery: The skill defines a robust process for discovering and verifying data sources. It includes guardrails that prioritize authoritative sources and require direct source verification before concluding analyses.
Delivery Routing Guardrails: The configuration includes logic to route dashboard delivery based on the environment (e.g., codex_desktop vs chatgpt_web), ensuring appropriate tool selection for different execution contexts.
Defined Security Boundaries: The instructions explicitly warn against bypassing source access limits and emphasize using modeled production tables over temporary or scratch tables, maintaining data integrity.
Artifact Safety: For MCP-based dashboards, the skill references shared mechanics (analytics-app-core.md) for payload safety, source provenance, and manifest validation before rendering.
Streamlit Development Practices: The specification for Streamlit dashboards encourages standard safety practices such as using st.cache_data for deterministic reads and explicitly managing secrets and environment variables for live deployments.

build-dashboard