uncertainty-imaging
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided JSON manifests to audit model reporting rigor, creating a potential surface for instructions embedded in the data.
- Ingestion point: The
scripts/check_uncertainty_reporting.pyscript reads external JSON files provided via the--manifestcommand-line argument. - Boundary markers: The script relies on standard
json.loads()for parsing structured data but does not include explicit warnings to the agent to ignore instructions embedded in the manifest values. - Capability inventory: The skill possesses file system access (
Read,Write,Edit) and shell execution (Bash) capabilities, which could be targeted by sophisticated injections in the manifest. - Sanitization: Content is validated for expected types (boolean, string, float) but is not sanitized against Natural Language instructions that might influence the agent's summary of the audit.
Audit Metadata