The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a surface for ingesting content from local documentation files, which could potentially contain malicious instructions.
Ingestion points: The skill instructions point to CONTEXT.md and CONTEXT-MAP.md as sources of information.
Boundary markers: There are no delimiters or specific instructions provided to the agent to treat the file content as data rather than instructions.
Capability inventory: The skill does not use any tools for network communication, file writing, or command execution, and it explicitly disables model invocation in its metadata.
Sanitization: No validation or sanitization of the documentation file content is performed.

wait-what