The Agent Skills Directory

[Prompt Injection] (SAFE): No attempts to override safety filters or bypass system instructions were detected. The task planning notes are standard procedural guidance.- [Indirect Prompt Injection] (LOW): The skill is designed to process conversation history, which is untrusted data. However, since the skill has no 'write' or 'execute' capabilities outside of summarizing its own context, the risk of an attacker influencing the system through this tool is negligible.- [Data Exposure] (SAFE): While the skill tracks active file paths and session summaries, it lacks any mechanism for data exfiltration or unauthorized access to sensitive files.- [No Code] (INFO): This is a markdown-only skill that provides instructions for the agent's behavior rather than executing code. There are no remote downloads or package dependencies.

compact