humanize-redo

Pass

Audited by Gen Agent Trust Hub on Jun 14, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by interpolating user-provided arguments ($ARGUMENTS) and reading potentially untrusted data from local files into its execution context.
  • Ingestion points: User input is ingested via the $ARGUMENTS variable, and existing workspace files (e.g., 01_input.txt, 02_detection.json) are read from the _workspace/ directory.
  • Boundary markers: The instructions do not use delimiters or explicit 'ignore embedded instructions' warnings when handling user input or file content.
  • Capability inventory: The agent is empowered to perform file system operations (globbing, writing, renaming) and call other specialized skills such as korean-style-rewriter and content-fidelity-auditor.
  • Sanitization: There is no evidence of validation or sanitization for the data processed from the workspace or the user-provided arguments.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 14, 2026, 04:31 AM
Security Audit — agent-trust-hub — humanize-redo