context-degradation
Pass
Audited by Gen Agent Trust Hub on Mar 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill primarily consists of instructional content and operational rules for the agent to monitor its own context window health. No malicious behavior, data exfiltration, or obfuscation was detected.
- [PROMPT_INJECTION]: The SKILL.md file contains directive language labeled as 'Iron Laws' and 'Mandatory Protocol'. These are internal performance constraints designed to guide the agent's logic for resource management and are not attempts to override safety filters.
- [COMMAND_EXECUTION]: The documentation defines a protocol for invoking auxiliary skills such as 'context-compressor' and 'session-handoff' as part of its corrective routing logic.
- [NO_CODE]: The main execution script (scripts/main.cjs) is currently a non-functional scaffold, posing no execution risk.
Audit Metadata