The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external data such as diagnosis summaries and user-provided actions which could influence the agent's behavior if malicious content is embedded in those inputs.
Ingestion points: Processes Diagnosis summary and Exact action requested as primary inputs to determine maintenance steps (SKILL.md).
Boundary markers: The skill includes a "Decision Tree" section that mandates explicit approval before touching production data or credentials, serving as a logical safety boundary.
Capability inventory: The skill uses the bruin CLI for operational commands (e.g., bruin cloud runs trigger, bruin run, bruin validate) and has the capability to modify repository files via pull requests (SKILL.md).
Sanitization: There are no specific instructions for sanitizing or escaping the diagnostic input before it is used to determine subsequent actions.

maintenance-action