context-window-hygiene
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill consists entirely of instructional Markdown content providing best practices for context engineering. It does not include any executable code, shell commands, or network-enabled tools.
- [NO_CODE]: No scripts (Python, Node.js), binaries, or package manifest files are present in the skill. The analysis confirmed the absence of any runtime execution logic.
- [INDIRECT_PROMPT_INJECTION]: While the skill discusses the processing of external data (specifically tool results), its primary guidance is to summarize and structure this data using typed schemas before it enters the conversation history. This practice acts as a defensive measure, reducing the risk of the agent misinterpreting large or malicious raw payloads embedded in tool outputs.
Audit Metadata