The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data by reading and searching through arbitrary source code files and configuration files in a project directory.
Ingestion points: Source code files in src/, data in assets/data/, and configuration files read using the Read, Glob, and Grep tools during Phases 2 and 3.
Boundary markers: The instructions lack explicit boundary markers or "ignore instructions" delimiters to differentiate between the agent's logic and the content of the files being audited.
Capability inventory: The skill uses the Bash, Write, and Task tools. A malicious file being audited could potentially contain instructions that attempt to trick the agent into executing arbitrary shell commands or writing unauthorized files.
Sanitization: There is no specified logic for sanitizing or escaping content read from files before it is passed to the security-engineer sub-task or processed by the primary agent.

security-audit