context-window-health-check
Pass
Audited by Gen Agent Trust Hub on May 15, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill operates entirely through natural language instructions within the LLM's chat interface. It does not request or utilize any external tools, network access, or file system operations.
- [NO_CODE]: The skill consists solely of Markdown documentation and prompt instructions. There are no executable scripts, binaries, or package management files (such as package.json or requirements.txt) included.
- [DATA_EXFILTRATION]: While the skill instructions direct the agent to analyze the current session's message count and topic diversity, it lacks any tools or commands (like curl or wget) to transmit this information to an external server.
- [PROMPT_INJECTION]: No malicious patterns such as safety filter bypasses, instruction overrides, or system prompt extraction techniques were detected. The skill includes explicit instructions to ignore unrelated user requests and maintain standard behavior.
Audit Metadata