MiniMax

PassAudited by ClawScan on May 1, 2026.

Overview

This instruction-only MiniMax skill is coherent and purpose-aligned, with disclosed use of a MiniMax API key, optional local memory, remote API/media workflows, and MCP boundaries that require user approval for sensitive actions.

This looks safe to install if you intend to use MiniMax APIs. Provide the MiniMax API key carefully, approve media uploads or paid jobs only after reviewing their scope and cost, and only enable local memory or remote MCP hosts when you understand what will be stored or shared.

Findings (4)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

If used, the agent may make MiniMax API calls using the user's account key, potentially consuming credits or accessing account-scoped resources.

Why it was flagged

The skill requires a MiniMax account credential, which is expected for direct MiniMax API calls but still grants access to account-backed provider usage.

Skill content
- `MINIMAX_API_KEY` for direct MiniMax API usage.
Recommendation

Use a dedicated or scoped MiniMax API key if available, keep it out of prompts and saved notes, and monitor provider usage.

What this means

These workflows could affect privacy, rights, cost, or remote tool behavior if approved without understanding the scope.

Why it was flagged

The skill covers high-impact actions such as media upload, voice imitation, remote tool use, and paid jobs, but it explicitly requires user approval before those actions.

Skill content
Explicit user approval before uploading private media, cloning or imitating a real person's voice, enabling remote MCP servers, or launching long-running paid generation jobs.
Recommendation

Before approving, confirm what media or voice data will be sent, what job may cost, and whether any remote tool has permission to read or change data.

What this means

Saved defaults, budget boundaries, consent notes, or MCP preferences may influence future MiniMax sessions.

Why it was flagged

The skill proposes persistent local memory for MiniMax preferences and workflow defaults; this is disclosed and consent-gated, but persistent notes can be reused across later tasks.

Skill content
Store data only in `~/minimax/` after user consent.
Recommendation

Allow persistence only if useful, avoid storing secrets or full private prompts/assets, and periodically review or delete files in ~/minimax/.

What this means

If remote MCP is enabled, external tools may receive data or perform actions depending on their scope.

Why it was flagged

The skill acknowledges that MCP servers are a trust boundary and requires approval before remote MCP use, which is appropriate for tool-mediated workflows.

Skill content
require explicit approval before enabling any remote MCP host
Recommendation

Approve only specific MCP hosts with clearly defined read/write permissions, and avoid broad remote MCP access unless necessary.