The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to interact with the local environment. These interactions are limited to invoking the official claude CLI for rule critique and git for repository file listing. Both are standard development tools and are used here for legitimate validation purposes.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill reads sensitive configuration files located in ~/.claude/settings.json and the project's .claude/ directory. This is the core functionality of the skill (managing these settings). Importantly, the skill does not perform any network operations to external or untrusted domains; all processing is local.
[UNVERIFIABLE_DEPENDENCIES]: The Python scripts included in the skill are designed to be standalone and use only the standard library. They declare no external dependencies in their PEP 723 script headers, significantly reducing the supply-chain attack surface.
[PRIVILEGE_ESCALATION]: The skill uses os.chmod to enforce restrictive file permissions (mode 0600) and directory permissions (mode 0700). This is a security best practice intended to protect the user's configuration files from unauthorized access by other users on the same host.
[INDIRECT_PROMPT_INJECTION]: The skill provides a natural language interface for editing settings via /automode-edit. While this creates an ingestion surface for untrusted data, the skill mitigates this by passing all proposed changes through a multi-stage validation pipeline: a JSON schema validator (which prevents the installation of persistent 'hooks'), a semantic linting tool, and a mandatory critique by the vendor's own CLI.
[PERSISTENCE]: The skill maintains project-specific state such as locks and backup files, but it does not attempt to modify system-level persistence mechanisms like shell profiles or cron jobs.
[RECOVERY_AND_INTEGRITY]: The skill implements a sophisticated 'repair' mechanism to handle interrupted writes or stale locks. It uses atomic file replacement and maintains a rolling history of backups with cryptographic hash labels to ensure data integrity and reversibility.

automode-config