Maestro Sdk

• Outcome: Review
• Latest audit: May 29, 2026 at 9:23 PM UTC
• Version:
  0.1.0

Overview

The skill is coherent as a Maestro Solana vault assistant, but it gives the agent broad financial authority with automatic key setup, persistent state, and transfer paths that need tighter user confirmation.

Review carefully before using with any funds. Use a dedicated limited wallet/session key, enforce strict Maestro vault policies, inspect the SDK package, and require explicit confirmation before every transfer or swap. Treat MEMORY.md and the memory logs as sensitive operational state, and clear or re-verify them when vault ownership or session keys change.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (4)

Vague Triggers

Medium

• Confidence: 91% confidence
• Finding: The trigger phrase "vault questions" is underspecified and can cause the skill to activate on loosely related conversation, leading to unsolicited vault discovery, health checks, or operational guidance. In a wallet-management skill, ambiguous activation increases the chance of unintended sensitive actions or stateful behavior in response to ordinary chat.

Vague Triggers

Medium

• Confidence: 95% confidence
• Finding: The category "Money/payment/balance" is extremely broad and likely to overlap with common conversational language, which can trigger vault discovery and transfer flows when the user did not intend to initiate financial operations. Because this skill controls a Solana vault and maps broad language directly to transaction execution paths, accidental activation could result in unauthorized or unintended transfer attempts.

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: Automatically generating or loading a private key and writing persistent state on skill load causes security-sensitive side effects before the user has granted informed consent. This creates risk of unapproved credential creation, filesystem modification, and persistence of wallet metadata, which is especially dangerous in a skill designed to control real vault funds.

Missing User Warnings

Low

• Confidence: 82% confidence
• Finding: Immediately disclosing the agent pubkey on load is a lower-severity issue, but it still exposes operational wallet metadata without user request or contextual warning. In combination with auto-init and persistence, this can leak identifying information, encourage premature funding, and normalize unsolicited operational prompts for a finance-related agent.

VirusTotal

66/66 vendors flagged this skill as clean.