The Agent Skills Directory

[SAFE]: The skill instructions and technical references provide standard architectural patterns for context management (compaction, isolation, etc.) with no evidence of prompt injection or safety bypass attempts.
[SAFE]: The Python module 'scripts/degradation_detector.py' uses legitimate libraries like 'numpy' and 're' to perform local text analysis and simulations, containing no network exfiltration, file system manipulation, or remote code execution patterns.
[SAFE]: While the skill contains fictional or futuristic model references (e.g., GPT-5.2) and benchmarks, these represent factual inaccuracies or placeholder content rather than a security threat such as metadata poisoning or deceptive behavior intended to compromise systems.

context-degradation