The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes local Python scripts measure.py and profile_store.py to perform text analysis and manage user settings. These scripts are self-contained and use only the Python standard library, with no network connectivity.
[CREDENTIALS_UNSAFE]: The profile management script profile_store.py contains an explicit blocklist (SENSITIVE_PARTS) to prevent the storage of secrets, tokens, passwords, or API keys in the user profile context.
[DATA_EXFILTRATION]: User preferences and personal writing baselines are stored locally in standard configuration directories (e.g., ~/.config/agent-skills/). No network operations were detected that would allow for data exfiltration.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text for auditing and humanization. The workflow mitigates common injection risks by recommending that long inputs be stored as temporary files rather than passed through direct shell arguments.
[PERSISTENCE]: The skill implements a persistence mechanism for user preferences (such as the default editing profile and baseline paths). This is the intended primary purpose of the skill's profile system and follows security best practices, such as setting restrictive file permissions (0o600) on generated configuration files.

lov-human-writing