The Agent Skills Directory

[SAFE]: The skill's primary function is to facilitate access to official public data repositories. The URLs provided in the 78 reference files target established government and academic institutions (e.g., nasa.gov, nih.gov, usgs.gov, worldbank.org). All external references fall within the scope of trusted organizations or well-known technology services.
[COMMAND_EXECUTION]: The skill instructions and reference files (e.g., sec-edgar.md, brenda.md) guide the agent to use platform-native tools like curl or generated Python scripts to interact with SOAP/REST APIs. This behavior is consistent with the skill's stated purpose of retrieving structured data from external sources.
[DATA_EXPOSURE]: The skill provides best-practice guidance for managing API keys, instructing the agent to look for credentials in environment variables or .env files rather than hardcoding them. No sensitive system paths are accessed beyond standard configuration files.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a surface for indirect prompt injection because it ingests raw JSON data from 78 external APIs. However, since these sources are verified public repositories and the risk is inherent to the data retrieval use case, the finding is assessed as low risk.
Ingestion points: External data enters the context from 78 scientific and government REST/SOAP APIs (e.g., pubchem.ncbi.nlm.nih.gov, api.nasa.gov).
Boundary markers: Absent. Instructions suggest presenting raw JSON results to the user.
Capability inventory: The agent possesses network capabilities (curl, WebFetch) and shell execution for Python scripts.
Sanitization: Not explicitly mandated in the instructions for the ingested API responses.

database-lookup