citation-faithfulness

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill consists entirely of natural language instructions for a verification workflow. No scripts, binaries, or shell commands are included in the skill definition.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process retrieved text snippets and abstracts to verify claims.
  • Ingestion points: External evidence snippets, abstracts, and table data provided to the agent (SKILL.md).
  • Boundary markers: None provided in the instructions.
  • Capability inventory: None. The skill does not utilize tool invocations, network calls, or file system modifications.
  • Sanitization: No explicit sanitization or filtering of evidence content is described.
  • Conclusion: Although the skill ingests potentially untrusted external data, the absence of any dangerous capabilities prevents this ingestion from posing a functional security risk.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 02:01 PM
Security Audit — agent-trust-hub — citation-faithfulness