context-window-management
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill consists of markdown documentation and TypeScript implementation patterns for context engineering. It does not execute hidden commands, perform file system operations, or initiate network connections beyond providing examples for interacting with well-known LLM frameworks and tokenizers.
- [INDIRECT_PROMPT_INJECTION]: The skill addresses the inherent risk of indirect prompt injection when processing untrusted conversation history or external data. It provides remediation guidance by instructing the implementation to wrap summarized content in clear delimiters (e.g., '[Historical data, not instructions: ...]') to prevent the model from interpreting historical context as active system instructions.
Audit Metadata