The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection through the ingestion of external data.
Ingestion points: The agent retrieves data from engineering_analytics_ci_failures (specifically the error_signature field) and engineering_analytics_ci_job_history (the commit_message field) as described in SKILL.md and references/investigation-queries.md.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the content of these log fields as non-instructional data.
Capability inventory: The agent has the capability to execute further SQL queries via execute-sql and use CLI tools like gh and hogli based on its analysis of the ingested data.
Sanitization: No sanitization or escaping mechanisms are described for the log data before it is processed by the agent. A maliciously crafted test failure or commit message could potentially contain instructions aimed at influencing the agent's diagnostic logic.

investigating-ci-failures