MiniMax Speech 2.8

Manage MiniMax Speech 2.8 TTS requests, voice catalog lookups, and precise voice/audio configuration using MiniMax API via CLI or script.

This appears to be a legitimate MiniMax TTS helper, but be cautious before installing or running it. Key points: - The script requires MINIMAX_API_KEY even though the registry metadata doesn't declare it; verify you are comfortable providing that API key and understand its privilege level (create a limited/test key if possible). - The CLI accepts an --endpoint override; do not point it at unknown or untrusted URLs, because your API key and audio payloads will be sent there. Prefer the default https://api.minimax.io endpoints or your organization's approved endpoints. - Inspect scripts/minimax_tts.py (already present) yourself — it is short and readable; run it in an isolated environment (virtualenv/container) and test with a limited-key account before using real/production credentials. - If you rely on the registry metadata for automation, update or correct the metadata to declare MINIMAX_API_KEY so tools and reviewers are not misled. If you want higher assurance, ask the publisher for a homepage/source repo, or request a signed/reproducible release from an identifiable maintainer.

Type: OpenClaw Skill Name: minimax-speech Version: 1.0.0 The `scripts/minimax_tts.py` skill, while intended for legitimate MiniMax API interaction, contains several vulnerabilities. The `--endpoint` argument allows the `MINIMAX_API_KEY` to be sent to an arbitrary URL, posing a risk of credential exfiltration if the AI agent is prompted to use a malicious endpoint. Additionally, the `--output` argument in both `tts` and `voices` subcommands is vulnerable to path traversal, potentially allowing arbitrary file writes outside the intended directory. The `decode_audio` function also allows downloading content from arbitrary URLs if the `output_format` is set to `url` and the API response contains a malicious URL. These are significant vulnerabilities that could be exploited via prompt injection or malicious input, but there is no clear evidence of intentional malicious design within the code or `SKILL.md`.