The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted external data, such as deployment histories, monitoring metrics, and health check responses, which creates a surface for indirect prompt injection.
Ingestion points: The scripts/get_previous_version.py script processes files and directories provided as input. The instructions in SKILL.md involve the agent monitoring health check responses and error rates from external sources.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish between data and instructions within the deployment logs or configuration files it processes.
Capability inventory: The skill performs file system write operations (shutil.copy2 and Path.mkdir) in scripts/rollback_deployment.py and scripts/verify_deployment.py. Additionally, the agent is granted access to powerful infrastructure tools including kubectl, docker, and git to execute the rollback procedures.
Sanitization: The provided scripts do not implement validation or sanitization of the input content before it is processed or used in reporting.

managing-deployment-rollbacks