The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches chemical, pharmacological, and safety data from the official PubChem API at pubchem.ncbi.nlm.nih.gov. This is a well-known scientific service provided by the National Center for Biotechnology Information (NCBI).
[COMMAND_EXECUTION]: Uses the uv tool to execute a local Python script (scripts/pubchem_api.py). This script acts as a controlled proxy for API interactions, enforcing rate limits and handling data parsing.
[DATA_EXPOSURE]: Query results are written to local JSON files on the filesystem. The skill instructions encourage the agent to read these files using tools like jq, following standard data processing workflows.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from an external API, creating a potential surface for indirect prompt injection.
Ingestion points: API responses are saved to local files (e.g., result.json) and then read by the agent.
Boundary markers: None are explicitly defined in the instructions for parsing the JSON output.
Capability inventory: The agent can execute the wrapper script and read/write files.
Sanitization: Input parameters for the API (like chemical names or SMILES strings) are properly URL-encoded using urllib.parse.quote to prevent path traversal or injection in the query itself.

pubchem-database