skills/zereight/skill/harness/Gen Agent Trust Hub

harness

Pass

Audited by Gen Agent Trust Hub on Jun 23, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: Analysis of the skill instructions and metadata revealed no indicators of malicious intent, obfuscation, or security policy violations.- [PROMPT_INJECTION]: The skill ingests user-provided text for 'goal' and 'rubric' to drive its improvement loop. This is an attack surface for indirect prompt injection where malicious instructions could be embedded in the data.
  • Ingestion points: harness_solve (parameters: goal, rubric) in SKILL.md.
  • Boundary markers: None present; the skill treats user-defined rubrics as authoritative criteria for evaluation.
  • Capability inventory: Executes a loop to generate content and writes resulting 'playbooks' to the local knowledge/ directory.
  • Sanitization: Not present; the tool directly utilizes user-defined strings to judge output quality.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 23, 2026, 05:14 AM
Security Audit — agent-trust-hub — harness