Security audit
Security checks for vulnerabilities and agentic risk
This MiniMax skill is a disclosed API workflow helper with local notes and remote generation calls that are scoped to the user's MiniMax tasks.
Install this only if you are comfortable sending prompts, approved media, and generation parameters to MiniMax services. Define activation and persistence preferences during setup, avoid storing full prompts or assets in ~/minimax/, and require explicit approval before private media uploads, voice imitation, paid long-running jobs, or remote MCP hosts.
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
integration: pending | done | declined
## Context
<!-- What the user is building with MiniMax and why -->
<!-- Example: Text plus speech workflow for narrated product demos with strict approval before media upload -->
## Activation Boundaries
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
- any rejected or unused assets that never get uploaded
This skill does NOT:
- treat compatible SDKs as exact feature matches without verification
- upload private media, voice references, or lyrics without explicit user intent
- enable remote MCP or broad tool access without explicit approval
- claim that every MiniMax modality is synchronous or instantly available
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
- confirm rights for images, music references, lyrics, logos, and character likenesses before upload
- treat long-running generation as paid remote compute and keep the user informed
- avoid hidden rerun loops that can burn credits without consent
The instruction says to use the file when the user says "which MiniMax model?" or when the workflow is "drifting," but the second condition is subjective and the quoted phrase could arise in general discussion rather than an explicit skill invocation. The file does not provide exclusion conditions or tighter scope to distinguish intended routing use from incidental mentions.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
## Common Mistakes
- Choosing a compatible interface first and only later picking the model.
- Using quality-first models in latency-critical loops without checking response budget.
- Treating video and music as if they will return like text.
- Forgetting that live model menus can change, which breaks stale pinned lists.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
## Common Mistakes
- Choosing a compatible interface first and only later picking the model.
- Using quality-first models in latency-critical loops without checking response budget.
- Treating video and music as if they will return like text.
- Forgetting that live model menus can change, which breaks stale pinned lists.
The activation guidance is overly broad, covering many related terms and use cases such as MiniMax, Hailuo, compatible SDKs, speech generation, and MCP-backed work. This can cause the skill to activate outside the user's intent, increasing the chance that it injects instructions or captures workflow context in unrelated conversations, especially because it is designed to act proactively.
No suspicious patterns detected.