principled-negotiation

Pass

Audited by Gen Agent Trust Hub on Jun 19, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements a purely reasoning-based workflow for negotiation preparation based on established methodologies (e.g., Harvard Program on Negotiation).
  • [SAFE]: Comprehensive safety instructions are included that explicitly prohibit the agent from requesting or disclosing sensitive information such as personal data, confidential contract terms, or private health facts.
  • [SAFE]: The skill maintains clear professional boundaries by instructing the agent to frame its output as preparation and recommend qualified legal or HR review for regulated matters.
  • [SAFE]: External references and grounding sources are restricted to trusted educational domains (pon.harvard.edu) and official documentation for established AI services (OpenAI, Anthropic).
  • [SAFE]: No malicious patterns such as prompt injection, code obfuscation, or remote code execution were found; static analysis flags for concealment appear to be false positives related to legitimate platform configuration (skill suppression) and the discussion of counterparty tactics.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 19, 2026, 04:35 AM
Security Audit — agent-trust-hub — principled-negotiation