The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill's documentation (specifically in references/skills-reference.md) provides examples of using the !command syntax to inject live data into a skill's context. Examples include fetching git information such as !git branch --show-current and !gh pr diff. These are standard patterns for gathering repository context and do not incorporate unvalidated user input into the shell commands.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform static analysis of the codebase. It runs commands such as ls, cat, and grep to identify project types (e.g., checking for package.json or pyproject.toml) and dependencies. These operations are limited to read-only discovery of the project structure.
[DATA_EXPOSURE]: The instructions include patterns for detecting sensitive files like .env or credentials.json. However, this is used specifically to recommend "Protection Hooks" that prevent the AI from accidentally editing or exposing these files, which is a defensive security recommendation.

Codex-automation-recommender