The Agent Skills Directory

[SAFE]: The skill explicitly promotes secure coding practices, such as instructing developers to use parameter binding for DuckDB SQL queries to prevent injection vulnerabilities and utilizing dlt.secrets.value for secure credential management rather than hardcoding sensitive information.
[COMMAND_EXECUTION]: The validation reference (references/validate-extension.md) instructs the user or agent to execute standard development tools (uv run pytest, ruff, mypy) within a temporary environment to ensure code quality and security. These commands are standard for development workflows and restricted to locally installed packages.
[INDIRECT_PROMPT_INJECTION]: The skill defines a framework for ingesting data from external REST APIs into a graph format, which introduces a potential surface for indirect prompt injection if the ingested data is later processed by an LLM.
Ingestion points: Data is collected from external API endpoints as defined in src/<pkg>/source.py.
Boundary markers: None explicitly defined in the templates, as the framework focuses on structured graph conversion rather than direct natural language processing of the output.
Capability inventory: The skill allows for network requests to external APIs and database operations using DuckDB.
Sanitization: The skill utilizes Pydantic models (BaseAsset) to enforce schema validation on all ingested resource data.

openhound