Mixlab Daily Briefing（ 高价值内容筛选）

• Outcome: Pass
• Latest audit: May 28, 2026 · 1:24 PM UTC
• Version: 3.2.7

Overview

This skill does what it claims: it fetches Mixdao items, uses an external LLM to create a briefing, saves local output, and uploads the generated briefing and recommendations back to Mixdao.

Install only if you intend the agent to update Mixdao briefing and recommendation records. Use scoped Mixdao and LLM API keys where possible, keep ANTHROPIC_BASE_URL pointed at a trusted endpoint, review generated content before running the upload step when accuracy matters, and delete temp files if fetched content should not remain on disk.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 86% confidence
• Finding: The skill uses sensitive environment variables and performs networked actions, but it does not declare permissions or clearly surface those capabilities in a machine-readable way. This is dangerous because users or hosting platforms may authorize or run it without realizing it can access secrets and call remote APIs, reducing informed consent and weakening policy enforcement.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 83% confidence
• Finding: The documented behavior does not fully match the operational behavior: it writes fetched and generated content to local temp files, may relax grouping constraints, and the loop semantics are ambiguous. Behavior mismatches are risky because operators may approve the skill under incomplete assumptions, while the actual implementation stores data locally and behaves differently from advertised, which can affect privacy, integrity, and auditability.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill performs remote modification by uploading generated briefings and recommendations back to mixdao, but the description does not prominently warn users that execution will change remote data. This is dangerous because a user may think the workflow is only analytical or local, when in fact it issues authenticated PATCH requests that can alter production content.

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: The script writes a Markdown briefing file to local disk automatically, derived from potentially sensitive fetched content, without any explicit user confirmation or disclosure at execution time. In an agent context, silent local persistence can create unexpected data retention and leak risk, especially on shared machines, CI runners, or synced working directories.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The generated briefing content is uploaded to an external mixdao API automatically, with no runtime confirmation or user-facing notice. Because the content is assembled from fetched items and LLM-generated summaries, this creates a real exfiltration boundary where potentially sensitive or proprietary material is transmitted off-host to a third party.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The script submits per-item recommendations to an external API in bulk without explicit user acknowledgement. In an autonomous or semi-autonomous agent setting, silently modifying remote records can cause unintended publication, integrity issues, or disclosure of model-generated content tied to internal identifiers.

VirusTotal

64/64 vendors flagged this skill as clean.