Security audit

Bread Protocal

Security checks across malware telemetry and agentic risk

Overview

The skill is coherent for Bread Protocol, but it needs review because it guides real wallet approvals and Base mainnet transactions without enough scoping or risk safeguards.

Install only if you specifically intend to use Bread Protocol. Do not paste a main-wallet private key into an agent; prefer a dedicated low-balance wallet or a wallet flow that requires manual signing. Before any transaction, independently verify getbread.fun, Base chain ID 8453, contract addresses, approval amounts, proposal IDs, ETH values, gas costs, and expected outcomes, and revoke allowances when finished.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (6)

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The trigger description is broad enough to activate on generic terms like 'wallet' or 'Bread', which can cause the skill to be invoked in situations unrelated to this protocol. In a skill that encourages wallet connection, token approvals, and ETH-backed on-chain actions, unintended invocation increases the chance a user is steered into risky financial workflows without sufficiently specific intent.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The skill instructs users to connect a wallet, fund with ETH, and approve BREAD spending without a prominent warning that approvals authorize token movement and that backing proposals commits real on-chain funds that may be illiquid, lost to fees, or exposed to contract risk. In this context, the omission is more dangerous because the skill is explicitly designed to drive financial transactions on a meme-coin launchpad, where users may act quickly and underestimate irreversible blockchain risks.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The examples instruct users to call approve() on an ERC-20 token and then send a payable backProposal() transaction on Base mainnet using real contract addresses, but they do not clearly warn that these operations authorize token spending and transfer real funds. In an agent skill context, copy-pasteable mainnet transaction examples can be acted on automatically or by inattentive users, increasing the risk of unintended fund loss or over-approval.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: This workflow includes real wallet signing and ERC-20 approval steps without an explicit warning that the action spends funds, grants token allowance, and creates irreversible on-chain effects. In an agent skill context, examples can be operationalized directly, so omitting a spending/risk notice increases the chance of unintended approvals or transactions by users who may treat the workflow as safe boilerplate.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The backing workflow instructs users to send ETH and approve BREAD fees without a clear upfront warning about financial loss, gas costs, volatility, and irreversibility. Because this skill is specifically for a meme-coin launchpad on Base, the surrounding context materially increases risk: users are being guided into speculative transfers where mistakes, phishing-style substitutions, or misunderstood examples can quickly cause real monetary loss.

Missing User Warnings

Low

Confidence: 89% confidence
Finding: The claim and refund sections trigger real on-chain write operations but do not explicitly warn that users will still pay gas and that transactions may fail or be economically irrational relative to claim value. While these actions are generally less dangerous than approvals or fresh ETH transfers, the omission can still mislead users into signing avoidable transactions or interacting under false assumptions about cost and reversibility.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.