The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill depends on the node-edge-tts npm package, which is an external library not maintained by a recognized trusted organization. It also references an external testing website at https://tts.travisvn.com/ for voice previewing.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text input to generate audio. It lacks explicit instructions or sanitization logic to handle text containing embedded instructions that might target the agent or downstream systems that process the generated audio.
Ingestion points: The text argument in tts-converter.js and the tts() tool call in SKILL.md accept raw user-provided strings.
Boundary markers: Absent. The instructions do not define delimiters or specify that the agent should ignore instructions embedded within the text to be converted.
Capability inventory: The skill possesses file write capabilities (via fs.writeFile in tts-converter.js and config-manager.js) and network access (to the Microsoft TTS service).
Sanitization: The logic in tts-converter.js only filters out specific trigger keywords (e.g., "tts") to prevent them from being spoken, but does not provide security-focused sanitization of the input text.
[COMMAND_EXECUTION]: The CLI tools tts-converter.js and config-manager.js expose parameters for defining file paths, such as --output and --config-path. If an agent uses these scripts with values derived directly from untrusted user input, it could lead to arbitrary file writes or the overwriting of sensitive configuration files.

edge-tts