two-tier-extraction

Pass

Audited by Gen Agent Trust Hub on Aug 29, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes large volumes of untrusted third-party content, such as email archives and document dumps, which creates a surface for indirect prompt injection attacks.
  • Ingestion points: Bulk content is ingested from external archives as described in SKILL.md.
  • Boundary markers: The skill implements a mechanism to flag detected imperatives (instructions) within the data using the untrusted_directives: true flag and dedicated markdown fences, ensuring they are not obeyed by the agent.
  • Capability inventory: The skill has permissions to write to various local directories (originals/, personal/, people/, companies/, sources/) and is marked as mutating: true.
  • Sanitization: The skill includes a 'Privacy Wall' that uses deterministic regex and exact-match rules to strip sensitive patterns (credentials, medical terms, legal terms) and private contacts from the data before it is sent to any LLM.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 29, 2026, 06:18 PM
Security Audit — agent-trust-hub — two-tier-extraction