before-after

Pass

Audited by Gen Agent Trust Hub on Jul 23, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill operates in a 'Chained Mode' where it ingests untrusted data from a local JSON file to populate its transformation dimensions. This creates a surface for indirect prompt injection if the source data contains malicious instructions intended to influence the agent's behavior during the mapping or drafting phase.
  • Ingestion points: Reads and extracts fields from ~/jtbd/<slug>/jtbd.json (specifically fields like switch_forces, problem, and needs).
  • Boundary markers: None; the skill does not wrap the ingested data in delimiters or provide 'ignore embedded instructions' warnings for the processing phase.
  • Capability inventory: The skill has permissions to read and write files to the local filesystem (~/jtbd/ directory) and can invoke image generation tools.
  • Sanitization: There is no evidence of sanitization, validation, or escaping of the external JSON content before it is interpolated into the drafting instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 23, 2026, 01:02 PM
Security Audit — agent-trust-hub — before-after