ask-exemplar

Pass

Audited by Gen Agent Trust Hub on Sep 22, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted data from the internet to perform its primary function of evaluating artifacts against external standards.
  • Ingestion points: The workflow requires searching or retrieving web pages to research external evidence for every invocation (SKILL.md).
  • Boundary markers: The skill requires source-grounded findings with links, classification labels, and confidence ratings to provide context and limit the influence of biased external data (SKILL.md).
  • Capability inventory: The skill is strictly limited to providing recommendations and is explicitly forbidden from editing files, committing, or pushing changes (SKILL.md).
  • Sanitization: The instructions include specific rules to ignore 'simulated advice from a famous person' and to prioritize primary official documentation over secondary sources (SKILL.md).
  • [SAFE]: The skill references documentation and best practices from recognized organizations including Anthropic, Apple, Google, Stripe, and Vercel as high-quality exemplars for technical research.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 22, 2026, 03:28 AM
Security Audit — agent-trust-hub — ask-exemplar