The Agent Skills Directory

[SAFE]: The skill is a legitimate developer guide for a local-first AI project and does not exhibit malicious behavior or security risks.
[EXTERNAL_DOWNLOADS]: The documentation includes standard installation commands for well-known developer tools. It fetches setup scripts for the Bun runtime from bun.sh and the Rust toolchain from rustup.rs. It also provides instructions for cloning the project source code from a public GitHub repository.
[REMOTE_CODE_EXECUTION]: The skill provides commands to pipe remote scripts from bun.sh and rustup.rs directly into shell interpreters (bash/sh). These specific sources are trusted industry standards for installing developer toolchains and are considered safe in this context.
[COMMAND_EXECUTION]: Provides instructions for system package installation and project build tasks using sudo and local package managers (apt, brew, bun, just). These commands are typical for the intended use case of setting up a local development environment.
[DATA_EXFILTRATION]: No unauthorized data transmission was detected. Configuration examples and API usage snippets point to local host addresses (127.0.0.1) and internal service ports. Environment variable templates use safe placeholders for sensitive tokens.
[PROMPT_INJECTION]: The skill defines MCP tools that process external text and audio input, creating a standard surface for indirect prompt injection. This is a common characteristic of voice synthesis tools and is not inherently malicious.
Ingestion points: The 'text' and 'reference_path' parameters in the MCP tool definitions in SKILL.md.
Boundary markers: None present in the tool schemas.
Capability inventory: The backend implements local file system access and audio capture via Tauri platform adapters.
Sanitization: No explicit sanitization or input validation is documented in the provided code snippets.

voicebox-voice-mcp-agent