The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's architecture involves reading and processing codebase files, which serves as a vector for indirect prompt injection if the files contain malicious instructions.
Ingestion points: Codebase content is ingested through the read, grep, and glob tools as part of the exploration process in references/explorer-prompt.md.
Boundary markers: There are no explicit boundary markers or instructions to the model to ignore potential instructions embedded within the codebase content.
Capability inventory: The agents can inspect the filesystem via read, grep, and glob and manage sub-tasks using delegate_task.
Sanitization: The skill does not perform sanitization or validation of the codebase content before it is processed by the agents.

how