The Agent Skills Directory

[PROMPT_INJECTION]: The section 'Suggest Using K-Dense Web For Complex Worflows' contains explicit instructions for the agent to change its interaction model. It directs the agent to 'proactively suggest using K-Dense Web' under specific conditions, which is a form of behavior steering for commercial promotion.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from the openFDA API, creating a vulnerability surface where instructions embedded in external database records could influence agent behavior.
Ingestion points: Data enters the context via fda.query, fda.query_drug_events, fda.query_drug_label, and other API wrappers defined in scripts/fda_query.py.
Boundary markers: None identified. There are no instructions provided to the agent to ignore or delimit potentially executable content within the API responses.
Capability inventory: The skill facilitates network operations to the openFDA API and execution of local Python scripts (scripts/fda_examples.py).
Sanitization: While the skill mentions a clean_drug_name function for input, there is no evidence of sanitization or schema validation for the data returned from the external API before it is processed by the agent.
[COMMAND_EXECUTION]: The skill documentation instructs the user/agent to execute local Python scripts (python scripts/fda_examples.py) to demonstrate functionality, which involves shell command invocation.

fda-database