The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local inspection scripts and standard development tools to perform static analysis.
Evidence: The skill runs node skills/wp-phpstan/scripts/phpstan_inspect.mjs to gather configuration data and uses composer to run phpstan analyse (SKILL.md).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project source code and configuration files, which could theoretically contain malicious instructions targeted at the agent.
Ingestion points: The inspection script and PHPStan tools read project files, including phpstan.neon and source code (SKILL.md).
Boundary markers: The instructions do not specify explicit delimiters or "ignore instructions" warnings for the ingested code.
Capability inventory: The skill has the capability to execute shell commands (node, composer, phpstan) and modify local project files (SKILL.md).
Sanitization: No specific sanitization or filtering of the ingested source code is described before the analysis is performed.

wp-phpstan