The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted source code, creating a surface for indirect prompt injection where malicious instructions in the audited code could attempt to manipulate the agent's reporting or behavior.
Ingestion points: Processes application source code files identified through directory traversal and pattern matching (**/auth/**, **/session/**, **/middleware/**) as described in SKILL.md.
Boundary markers: Lacks explicit instructions or delimiters to treat the analyzed code content as potentially hostile data to avoid unintended instruction following.
Capability inventory: The skill employs Bash, Write, and Edit tools to perform security audits and generate report files in the security-reports/ directory.
Sanitization: No evidence of sanitization or validation of the ingested source code content is provided before it is processed or included in output reports.

checking-session-security