relsa-severity-assessment

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external CSV and TSV files, which creates a potential surface for indirect prompt injection attacks.
  • Ingestion points: The read_relsa_table and canonicalize functions in scripts/_common.py load user-provided data into the agent's context.
  • Boundary markers: The skill does not employ specific delimiters or clear instructional boundaries to separate the processed data from the agent's operating context.
  • Capability inventory: The skill possesses significant local capabilities, including reading and writing files to the disk and executing complex statistical modeling scripts.
  • Sanitization: While measurement columns are validated using pd.to_numeric, descriptive metadata columns (such as id, treatment, and condition) are processed as raw strings, allowing embedded instructions to remain intact.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 04:56 AM
Security Audit — agent-trust-hub — relsa-severity-assessment