persuasion-cialdini-influence-design

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • Prompt Injection (SAFE): No patterns detected that attempt to override system instructions or bypass safety filters. The instructions focus on structured output and ethical constraints.
  • Data Exposure & Exfiltration (SAFE): No access to sensitive local files, hardcoded credentials, or external network requests were found. All references are local markdown files.
  • Unverifiable Dependencies & Remote Code Execution (SAFE): The skill consists entirely of markdown files. It does not install packages or execute scripts (Python, Node.js, or Shell).
  • Indirect Prompt Injection (LOW): The skill processes untrusted user input ($ARGUMENTS) to generate persuasive content.
  • Ingestion points: $ARGUMENTS variable in SKILL.md.
  • Boundary markers: Absent; inputs are not explicitly delimited or escaped.
  • Capability inventory: Text generation only; no subprocess calls or network operations.
  • Sanitization: Includes a mandatory 'ethical risk check' and a 'traceability matrix' (principle -> evidence -> claim) to ensure output is grounded in provided facts rather than hallucinated or injected instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:09 PM