The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting and analyzing arbitrary PHP source code from a user-specified directory, which introduces a potential attack surface for indirect prompt injection. Maliciously crafted content within the target project's files could attempt to manipulate the agent's auditing logic.
Ingestion points: The skill instructions specify recursive traversal and reading of all PHP source files located within the user-provided source_path.
Boundary markers: There are no explicit instructions to use delimiters or ignore-embedded-instruction markers when reading file content, which increases the risk that code comments or data within the source files could be interpreted as part of the agent's operating instructions.
Capability inventory: The skill is equipped to perform deep directory traversal, read file content, and orchestrate the execution of a multi-stage auditing pipeline involving several other specialized security tools.
Sanitization: The workflow does not include a step for sanitizing or escaping the content of the analyzed files before they are processed by the core logic.

php-audit-pipeline