The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection (Category 8) because its primary function is to ingest and analyze untrusted PHP code and HTTP parameters. It lacks explicit boundary markers or instructions to sanitize these inputs before they are processed by the agent or used to generate probes.
Ingestion points: PHP source code, HTTP request parameters ($_GET, $_POST), and JSON objects provided for analysis (SKILL.md).
Boundary markers: Absent.
Capability inventory: Shell command execution via php -r (SKILL.md).
Sanitization: Absent.
[COMMAND_EXECUTION]: The skill instructs the agent to perform dynamic execution (Category 10) by running local PHP probes via the command line interface. Specifically, it suggests using php -r to execute code snippets for verifying type juggling behavior. This is categorized as low risk as it relies on simple script generation from known templates for debugging purposes.

type-juggling