The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted metadata from the public npm registry, which could contain malicious instructions designed to influence the agent's behavior during the translation or presentation phase.
Ingestion points: Package names and descriptions are retrieved from the registry.npmjs.org search API within scripts/pi_trending.py.
Boundary markers: The skill lacks explicit delimiters or instructions for the agent to treat the external metadata as untrusted data, increasing the risk that the agent might follow instructions embedded in a package description.
Capability inventory: The agent is specifically instructed to translate and summarize package descriptions. Furthermore, SKILL.md provides command templates for installing discovered packages, creating a surface where malicious package names could influence agent actions.
Sanitization: The Python script does not perform any sanitization or filtering of the description or name fields returned by the npm registry API before they are presented to the agent.

pi-trending