The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's automated workflow creates an attack surface for indirect prompt injection by ingesting and acting upon untrusted data from the local codebase.
Ingestion points: The fuse-ai-pilot:explore-codebase agent call defined in SKILL.md is instructed to detect and read existing style configurations (.php-cs-fixer.php, phpcs.xml) and composer.json files within the user's project.
Boundary markers: The instructions do not define explicit boundary markers or provide specific guidance to sub-agents on how to distinguish between legitimate configuration data and potential malicious instructions embedded within those files.
Capability inventory: The skill possesses the capability to modify critical project files (PHP source files, composer.json, and directory structures) and execute shell-level commands via the Composer toolchain (composer check, composer require, etc.).
Sanitization: There is no explicit requirement for the agent to sanitize or validate the content of the detected configuration files before using them to drive subsequent file modifications or tool executions.

php-standards