The Agent Skills Directory

[SAFE]: No malicious patterns, prompt injections, or unauthorized data access were detected. The skill instructions focus on establishing a rigorous framework for task execution and verification.
[SAFE]: The skill uses external verifiers (e.g., exit codes, logs) as primary signals for success, which is a defensive design against 'hallucination' or self-evaluation bias.
[SAFE]: All external references point to reputable research institutions (Anthropic, Microsoft Research, arXiv) or well-known development platforms (GitHub) and are used solely for documenting design rationale.

goal-loop