dialogue-audio
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references an external installation guide hosted on a remote GitHub repository at
https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md. - [EXTERNAL_DOWNLOADS]: The skill recommends installing multiple external skills (e.g.,
belt-sh/cli,inference-sh/skills@elevenlabs-dialogue) which fetches code from external GitHub organizations. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text data from user prompts and interpolates it into JSON payloads that are executed via the
beltCLI tool to interact with thefalai/dia-ttsremote application. - Ingestion points: Multiple examples in
SKILL.mdshow user-controllable text being placed into thepromptfield of a JSON object. - Boundary markers: The prompt content is encapsulated within a JSON string, which provides structural separation but lacks explicit instructions to the AI agent to ignore embedded commands or malicious text.
- Capability inventory: The skill uses the
beltCLI tool to execute remote audio generation and media merging applications. - Sanitization: There is no evidence of input validation, escaping, or filtering of the user-provided prompt content before it is processed by the remote applications.
Audit Metadata