evidence-ledger-research

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [SAFE]: No malicious patterns, obfuscation, or data exfiltration attempts were detected in the skill contents.
  • [NO_CODE]: The skill is composed exclusively of natural language instructions and YAML configuration files, with no provided scripts or binaries.
  • [PROMPT_INJECTION]: The skill is designed to process untrusted data including documents, URLs, and data tables (ingestion points). While this creates a surface for indirect prompt injection, the skill lacks high-risk capabilities like arbitrary command execution or file system writes, and it employs boundary-like structures via the mandatory 'Evidence Ledger' and 'Source Discipline' instructions to maintain grounding.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 03:33 PM
Security Audit — agent-trust-hub — evidence-ledger-research