The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands via the bash tool provided by an external MCP. It specifically uses the fda utility to perform its primary function of querying regulatory records.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external FDA records which creates a surface for indirect prompt injection if the retrieved content contains adversarial instructions.
Ingestion points: Output from the fda command executed via the bash tool as described in SKILL.md.
Boundary markers: The instructions lack specific delimiters or warnings to the agent to ignore instructions that might be embedded within the external record data.
Capability inventory: The agent has access to bash (described as a read-only shell) and ask tools, which provides a capability surface for executing commands or interacting with the user.
Sanitization: There are no instructions provided to sanitize, escape, or validate the content retrieved from external records before it is processed by the agent.

fda-lookup