The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash, Read, and Glob tools to scan local files (e.g., CLAUDE.md, .cursorrules, .github/copilot-instructions.md) for specific text patterns related to the AIWG framework.
[DATA_EXPOSURE_SURFACE]: The skill accesses several project-level configuration files to identify if context hooks are enabled. This is limited to local read-only inspection of text patterns and does not involve accessing sensitive system credentials or network communication.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project files as part of its detection logic.
Ingestion points: Files like CLAUDE.md, AGENTS.md, and .cursorrules are read from the local repository.
Boundary markers: None explicitly defined in the provided workflow, though the skill's logic is restricted to pattern matching for specific directives (e.g., @AIWG.md).
Capability inventory: Uses Bash, Read, and Glob for file system inspection.
Sanitization: None described, as the skill only reports on the existence of specific substrings and metadata (like file size) rather than executing or interpreting the content of the files.

hook-status