The Agent Skills Directory

[Prompt Injection] (SAFE): The text contains instructional guidance on session management and does not include patterns aimed at bypassing LLM safety guardrails or system prompts.- [Data Exposure & Exfiltration] (SAFE): No network calls, credential patterns, or exfiltration vectors were detected. Recommendations to read persistent memory are consistent with standard agentic state management.- [Remote Code Execution] (SAFE): No remote script downloads or dynamic code execution patterns are present in the provided file.- [Indirect Prompt Injection] (SAFE): The skill identifies ingestion points like memory files and handoff notes as part of its workflow. While it lacks explicit sanitization instructions for these sources, it is primarily a conceptual framework rather than a programmatic implementation, posing no immediate risk.- [Persistence Mechanisms] (SAFE): References to persistent context relate to maintaining session state through standard agent memory files rather than unauthorized system-level persistence like crontabs or shell profile modifications.

session-lifecycle