skills/tt-a1i/hermes-agent/humanizer/Gen Agent Trust Hub

humanizer

Pass

Audited by Gen Agent Trust Hub on Aug 28, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting external text from user prompts or local files to identify and rewrite AI-isms, creating an attack surface for instructions embedded in the processed data.
  • Ingestion points: Untrusted data is ingested through inline text provided by the user or by reading local file paths as described in the 'How to use it' section of SKILL.md.
  • Capability inventory: The skill instructs the agent to use file-system tools including read_file, patch, and write_file to process and store the humanized text.
  • Boundary markers: The instructions lack the use of delimiters or explicit 'ignore embedded instructions' warnings for the data being humanized.
  • Sanitization: There are no defined procedures for sanitizing, escaping, or validating the input text before it is processed by the agent and written back to the filesystem.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 28, 2026, 12:48 PM
Security Audit — agent-trust-hub — humanizer