The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill is highly vulnerable to indirect prompt injection attacks because it combines data retrieval from an external source with high-privilege administrative capabilities.
Ingestion points: The index.query() function in SKILL.md and references/deployment.md retrieves metadata and values from the Pinecone database.
Boundary markers: Absent. There are no instructions or delimiters provided to help the agent distinguish between internal system instructions and retrieved data.
Capability inventory: The skill includes destructive operations such as pc.delete_index() and index.delete() in SKILL.md.
Sanitization: Absent. The documentation does not implement or suggest any filtering or sanitization of retrieved metadata before it is used in the agent's context.
[External Downloads] (MEDIUM): The skill requires the installation of external software from an untrusted source.
Evidence: pip install pinecone-client and dependencies: [pinecone-client] in SKILL.md.
Risk: The package pinecone-client is not hosted by a specifically whitelisted trusted organization, creating a potential risk for supply chain attacks or dependency confusion if versions are not pinned and verified.

pinecone