The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill represents an attack surface for indirect prompt injection because it fetches and processes content from an external, community-contributed source (protocols.io).
Ingestion points: The agent retrieves data from multiple endpoints including GET /protocols/{id} (protocol content/steps), GET /protocols/{id}/comments (user-generated feedback), and GET /workspaces/{id}/files/search (file metadata and descriptions).
Boundary markers: The skill instructions do not explicitly mandate the use of delimiters (like XML tags or triple quotes) or provide specific instructions for the agent to ignore potentially malicious instructions embedded within the retrieved scientific data.
Capability inventory: The skill provides the agent with capabilities to perform network operations (via requests to the protocols.io API) and handle local file operations (suggested in Python examples for uploading data).
Sanitization: There is no explicit guidance within the skill's reference files regarding the sanitization or validation of the content fetched from the API before it is processed by the agent.

protocolsio-integration