The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to utilize various command-line tools to verify language signatures and explore project structures. In SKILL.md, the agent is directed to use php -r for syntax verification. In READING-A-CODEBASE.md, it is encouraged to run project-local scripts such as php artisan route:list for Laravel or php bin/console debug:router for Symfony. In SYNTAX-8X.md, the agent is prompted to run php -v and composer show to identify environment constraints.
[DYNAMIC_EXECUTION]: The agent is prompted to generate and execute PHP code snippets locally using the command line to verify facts about the language. The primary evidence is the instruction to use php -r '...' in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes untrusted user-supplied codebases while retaining instructions to execute scripts found within those directories. * Ingestion points: User-provided PHP files, folders, and repositories described in SKILL.md and READING-A-CODEBASE.md. * Boundary markers: Absent. The instructions do not define specific delimiters or security headers to isolate user code from the agent's logic. * Capability inventory: Execution of php (via php -r), composer, and framework-specific binaries (artisan, console) located in the user's project root. * Sanitization: Absent. The skill does not provide logic for the agent to verify the safety or authenticity of the artisan or console binaries before execution.

php-mentor