The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute high-risk commands, including rm -rf for directory deletion and osascript to quit running applications. These are core functions of the machine cleanup process but provide a mechanism for significant system modification.
[CREDENTIALS_UNSAFE]: The scripts/inventory.sh script and the dotfolder-scanner-agent perform a comprehensive audit of sensitive system paths, including SSH private keys (~/.ssh/id_rsa, ~/.ssh/id_ed25519), GPG keys (~/.gnupg), and cloud provider credentials (~/.aws/credentials, ~/.config/gcloud/credentials.db). This data is accessed to identify "auth-bearing" folders and prevent accidental deletion, but it exposes a wide range of secrets to the agent's context during the survey.
[PROMPT_INJECTION]: The skill represents an attack surface for indirect prompt injection by ingesting local machine state—such as filenames, process lists, and shell configuration lines—into the agent's context without explicit technical sanitization or boundary markers.
Ingestion points: scripts/inventory.sh collects file system metadata and authentication hints from paths throughout the home directory. package-inventory-agent captures global package lists from multiple managers.
Boundary markers: The skill relies on natural language instructions (e.g., "6 Golden Rules") and a separate "Critic Agent" for validation, but lacks technical delimiters for all ingested data.
Capability inventory: The skill possesses high-privilege capabilities via the Bash tool, specifically for file deletion (rm -rf) and process control (osascript, pgrep).
Sanitization: The skill uses regex-based classification for files but does not describe specific escaping of file system data before interpolation into prompts.

clean-machine