voice-dna-kit
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local shell script
scripts/analyze-style.shwhich wraps a Python script for text analysis. - Evidence: The workflow in
SKILL.md(Step 3) explicitly instructs the agent to resolve and runscripts/analyze-style.shagainst cleaned samples. - The script performs benign statistical measurements (word counts, punctuation frequency, lexical diversity) and does not interact with the network or sensitive system files.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted text samples from external authors, creating a surface for indirect prompt injection.
- Ingestion points:
SKILL.md(Step 1) andscripts/analyze-style.shprocess three to five writing samples provided by the user. - Boundary markers: The skill lacks formal boundary delimiters for user samples, though it instructs the agent to "Exclude quoted third-party text" and "Remove navigation, metadata, signatures, code, URLs".
- Capability inventory: The agent has the capability to write multiple files (
voice-dna.md,style-dna.md, etc.) and execute a local script. - Sanitization:
scripts/analyze-style.shincludes aclean_markdownfunction that removes code blocks, links, and headers before analysis, reducing the risk of the agent interpreting control characters in the samples.
Audit Metadata