brand-voice-governance

Pass

Audited by Gen Agent Trust Hub on Aug 3, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [SAFE]: A comprehensive analysis of the skill's instructions and the associated Node.js script found no malicious patterns, prompt injections, or obfuscation techniques.
  • [COMMAND_EXECUTION]: The skill executes a local script, voice-check.js, to perform text analysis. The script is dependency-free, does not perform network requests, and handles input through standard JSON parsing without the use of dangerous functions like eval().
  • [DATA_EXPOSURE]: The skill accesses local marketing documentation (e.g., .agents/product-marketing.md) to establish the voice profile. This data usage is confined to the local execution environment for the purpose of governance auditing and does not involve exfiltration to external domains.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input text and third-party brand briefs to perform its audit.
  • Ingestion points: User-provided text and brand briefs are ingested via JSON arguments to the voice-check.js script.
  • Boundary markers: The script processes input by splitting it into discrete text blocks and matching against literal strings or regex patterns, reducing the likelihood of instruction-based overrides within the data.
  • Capability inventory: The skill uses node to run local analysis; it lacks the ability to write to sensitive system files or access the network.
  • Sanitization: Inputs are parsed as standard JSON, and text analysis is performed using literal string matching and established heuristic formulas (Flesch–Kincaid grade levels).
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 3, 2026, 05:04 PM
Security Audit — agent-trust-hub — brand-voice-governance