long-horizon-prompting

Pass

Audited by Gen Agent Trust Hub on Jul 14, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill consists of technical documentation and prompt templates; it contains no executable scripts, shell commands, or unauthorized tool invocations.
  • [SAFE]: All external links point to reputable sources including official vendor documentation from OpenAI and Anthropic, the arXiv research repository, and the METR evaluation platform.
  • [SAFE]: The content promotes security and reliability best practices for agent development, such as modular verification, evidence-grounded progress reporting, and the use of fresh-context adversarial auditors.
  • [SAFE]: No evidence of obfuscation, data exfiltration, or persistence mechanisms was found in the text or metadata.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 14, 2026, 05:57 PM
Security Audit — agent-trust-hub — long-horizon-prompting