The Agent Skills Directory

[COMMAND_EXECUTION]: The skill operates by executing a local Python script scripts/deep_research.py with various subcommands (start, poll, result, etc.). These commands incorporate user-supplied research themes as arguments, which is a standard pattern for this type of tool.
[EXTERNAL_DOWNLOADS]: The documentation instructs the user to install the openai and google-genai Python packages from standard registries. These are official libraries for the referenced well-known services.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks. It retrieves data from external web sources via the research APIs and saves this content to markdown files (e.g., in docs/research/). If the agent later reads or processes these research reports, malicious instructions embedded in the external web content could potentially influence the agent's behavior.
Ingestion points: Research results fetched from OpenAI/Google APIs via python scripts/deep_research.py result.
Boundary markers: Not explicitly defined in the provided instructions for output processing.
Capability inventory: The skill performs file writes to the local filesystem (docs/research/) and requires network access to communicate with API providers.
Sanitization: No specific sanitization or filtering of the retrieved research content is mentioned in the documentation.

deep-research