ai-social-media-content

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill workflows interpolate user-defined variables and AI-generated text into subsequent commands, which could be exploited if malicious content is introduced into the chain.
  • Ingestion points: Shell variables defined by the user (e.g., CONCEPT, TOPICS) and generated script outputs from the 'openrouter/claude-sonnet-45' model are used as inputs for later content generation steps.
  • Boundary markers: Data is passed as string values within JSON objects via the --input flag of the belt CLI, providing basic structural separation.
  • Capability inventory: The skill utilizes the Bash tool (restricted to the belt command) to interact with remote AI inference services and social media platforms (Twitter/X).
  • Sanitization: No explicit validation, escaping, or filtering of the variable content or generated scripts is performed before they are processed by subsequent tool calls.
  • [EXTERNAL_DOWNLOADS]: The skill facilitates the installation of additional capabilities from the developer's official repositories.
  • Evidence: The skill instructs users to install belt-sh/cli and various inference-sh/skills using the 'npx skills add' command, which fetches assets from the vendor's GitHub-hosted infrastructure.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 08:14 PM
Security Audit — agent-trust-hub — ai-social-media-content