The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes processes that ingest data from external sources, creating a potential attack surface. \n- Ingestion points: External data is accessed from database tables using psql in SKILL.md and references/API_REFERENCE.md, and from Guidewire Cloud Console APIs using curl in SKILL.md and references/API_REFERENCE.md. \n- Boundary markers: Absent; no explicit markers are defined to separate ingested data from agent instructions. \n- Capability inventory: The skill leverages tools such as Bash (command execution), curl (network operations), aws (file-write and network operations), psql (database access), and gh (network operations) across all referenced files. \n- Sanitization: Absent; the provided examples do not include explicit sanitization of data retrieved from external sources before it is processed. \n- [DYNAMIC_EXECUTION]: The skill includes a pattern for dynamically loading environment variables. \n- Evidence: The use of eval to export shell variables decrypted by sops in the API_REFERENCE.md workflow example. \n- Context: This is a standard secret management practice used to securely inject environment variables and does not involve untrusted input in the context of this skill.

guidewire-ci-cd-pipeline