The Agent Skills Directory

[DATA_EXFILTRATION]: The skill manages a local user profile file (typically in ~/.config/agent-skills/profile.json) using the scripts/profile_store.py utility. This script includes a safety mechanism to prevent the persistence of sensitive keys such as tokens, passwords, and API keys.
[COMMAND_EXECUTION]: The skill uses local Python scripts for specific tasks: scripts/profile_store.py is used to read and write user preferences and brand identity facts, while scripts/validate_skill.py is used for linting and validating the skill structure. These scripts operate within standard user permissions.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided project descriptions to generate naming recommendations. It has the following profile regarding indirect injection risks:
Ingestion points: Project purpose, audience, and language requirements are parsed from the natural language request (SKILL.md).
Boundary markers: The execution boundary instructions explicitly direct the agent to parse specific project facts and protect existing uncommitted content.
Capability inventory: The skill can perform file writes to its local profile via scripts/profile_store.py.
Sanitization: The profile_store.py script filters for secret-like keywords (token, password, api_key, etc.) and requires a --confirm flag for writing data, ensuring user oversight.

lov-gen-project-name