The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill manages a workflow that involves reading and processing external data which could potentially contain malicious instructions intended to influence the agent's behavior. Ingestion points: The skill is designed to ingest data from error logs, project configurations, and source code history (Step 1, Step 3, and Step 4 in SKILL.md). Boundary markers: There are no specific boundary markers or instructions to the agent to disregard natural language instructions found within the logs or code being investigated. Capability inventory: The skill is granted access to tools including Bash, Read, and Edit in the YAML frontmatter, which could be exploited if the agent follows instructions embedded in external data. Sanitization: While the skill lacks specific sanitization for indirect prompt injection, it includes a defensive instruction to redact secrets and prefer presence or equality checks for sensitive values when comparing configurations (Step 3 in SKILL.md).

investigate