The Agent Skills Directory

[Prompt Injection] (SAFE): The instructions in SKILL.md and README.md provide clear, benign process-oriented guidance for documentation tasks. No attempts to override agent safety filters or bypass system prompts were detected.
[Data Exposure & Exfiltration] (SAFE): The 'internal.md' template includes placeholders for documenting environment variables like API_KEY and DATABASE_URL. These are strictly for user documentation purposes and are not associated with any network commands or data collection logic.
[Unverifiable Dependencies & Remote Code Execution] (SAFE): This skill contains no executable code (Python, JavaScript, or Shell scripts) and no package management files. It functions entirely through Markdown templates and instructions.
[Indirect Prompt Injection] (LOW): The skill involves reading project files (e.g., package.json) to verify documentation accuracy. This represents a standard documentation use case where the agent processes untrusted local data, but the skill lacks the capabilities (like network access or command execution) to be exploited via this vector.

crafting-effective-readmes