voice-calibration

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided writing samples, conversation history, and external exemplars to generate a durable voice profile.
  • Ingestion points: Processes content from user-approved folders, conversation history, and provided examples as described in SKILL.md (Sections 1, 2, and 5).
  • Boundary markers: The skill requires an explicit 'approval scope' and instructs the agent to disregard unapproved or sensitive content.
  • Capability inventory: Writing to the filesystem (templates/voice-profile.md), resolving companion skills (writing-core), and recommending CLI command execution.
  • Sanitization: Instructs the agent to extract structural writing principles rather than copying raw text and explicitly forbids storing sensitive samples in the resulting profile.
  • [EXTERNAL_DOWNLOADS]: Suggests the installation of a companion skill, writing-core, from the author's own repository (George-RD/growth-arsenal) using the npx skills add command. This represents standard modular dependency management for the vendor's toolset.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 01:48 PM
Security Audit — agent-trust-hub — voice-calibration