The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to create, modify, and execute shell scripts (hooks) that run automatically in response to agent events. This includes making scripts executable using chmod +x as shown in the workflow section of SKILL.md.
[CREDENTIALS_UNSAFE]: The skill facilitates access to sensitive files, including the agent's main configuration (~/.claude/settings.json) and environment files (.claude-env, .env). Templates provided in references/claude-templates.md demonstrate patterns for monitoring access to .env, secrets, and credentials files.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it reads configuration data from .claude/settings.json within project directories, which could be controlled by a third party in a malicious repository.
Ingestion points: The skill reads ~/.claude/settings.json, .claude/settings.json, and .claude/settings.local.json (documented in SKILL.md).
Boundary markers: No delimiters or safety instructions are used when reading or processing these configuration files to prevent the agent from following instructions embedded within the data.
Capability inventory: The skill provides full shell execution capabilities via the hook system, file writing access, and the ability to change file permissions (SKILL.md).
Sanitization: While the skill includes scripts/validate_hooks.py to check JSON structure, it does not perform security-focused sanitization of the shell commands configured within the hooks.

hooks-management