social-kit

Pass

Audited by Gen Agent Trust Hub on Jul 2, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill metadata specifies installation via npx goose-skills, which downloads the necessary components from the vendor's official distribution channels. This is an expected part of the skill's setup process.
  • [DATA_EXPOSURE]: The skill reads configuration data and voice guide templates from the ~/.goose-skills/ directory. This access is localized to the skill's own operational data and is used to personalize the content generation process.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided content via the --brief argument and passes it to downstream content generation tools.
  • Ingestion points: User-provided text in the --brief flag.
  • Boundary markers: The input is passed as a quoted string to sub-skills without additional isolation or instructions to ignore embedded directives.
  • Capability inventory: The skill orchestrates multiple sub-tools and performs file system operations to organize the generated content.
  • Sanitization: No validation or sanitization of the brief is performed before it is utilized in subsequent tool calls.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 2, 2026, 06:18 AM
Security Audit — agent-trust-hub — social-kit