The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the belt CLI tool to execute various AI models for text-to-speech, voice cloning, and media merging. Access to this tool is explicitly defined in the allowed-tools frontmatter.
[EXTERNAL_DOWNLOADS]: The documentation instructs the user to install external dependencies using npx skills add belt-sh/cli and references installation guides hosted on GitHub (raw.githubusercontent.com/inference-sh/skills/...). These resources are associated with the platform's standard ecosystem.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests untrusted text data from the user and passes it directly to remote inference APIs through the belt CLI.
Ingestion points: User-supplied prompt and text fields in the JSON payloads processed in SKILL.md.
Boundary markers: None are present in the provided examples to delimit user content from system instructions.
Capability inventory: Execution of shell commands via the belt tool and potential file system writes using redirection (e.g., > speaker1.json).
Sanitization: There is no evidence of input validation or escaping for the text content before it is passed to the CLI.

ai-voice-cloning