The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze potentially untrusted data from external sources, which creates a surface for indirect prompt injection.
Ingestion points: In SKILL.md, the agent is instructed to "Capture the reported symptom" and "Read logs/code/tests." The execution-protocol.md further specifies gathering "Error messages," "Reproduction steps," "logs," and "environment notes."
Boundary markers: The skill relies on a general execution-policy.md for authorization but does not define specific prompt delimiters or "ignore embedded instructions" warnings for the data processed during the "PREPARE" and "ACQUIRE" scenes.
Capability inventory: The agent has significant capabilities including writing code patches (WRITE in SKILL.md), executing reproduction commands (CALL_TOOL in SKILL.md), and running project verification checks (oma agent verify in result-contract.md).
Sanitization: There is no evidence of input validation or sanitization applied to logs or error messages before they are used to "Infer root cause" or "Write minimal fix."

oma-debug