The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data from external sources, creating a potential surface for indirect prompt injection attacks. * Ingestion points: The Express.js routes in src/index.js and the Cloud Run Functions in SKILL.md ingest data via HTTP request parameters, body, and query strings, as well as Pub/Sub message payloads. * Boundary markers: There are no explicit boundary markers or instructions to the agent to disregard embedded commands in the processed data. * Capability inventory: The skill utilizes powerful deployment tools including gcloud run deploy, gcloud functions deploy, and docker for image creation and management. * Sanitization: The provided code snippets use standard JSON.parse and basic parameter extraction but lack application-level validation or escaping mechanisms to prevent malicious content from influencing future agent actions.

gcp-cloud-run