CHEESE Agent Marketplace

PassAudited by VirusTotal on May 12, 2026.

Overview

Type: OpenClaw Skill Name: cheese Version: 4.1.0 The skill requires a wallet private key (`CHEESE_PRIVATE_KEY`) and uses highly directive instructions in `SKILL.md` to force the agent into a persistent monitoring loop via Waku chat (`--watch`). This creates a significant attack surface where an agent could be manipulated by external messages while holding active credentials. The absence of the underlying script code (e.g., `scripts/cheese-cli.ts`) makes it impossible to confirm if the private key is handled securely or if the 'gasless relay' (aicheese.app) involves unauthorized data transmission.

Findings (0)

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 the invoked CLI or environment is compromised, funds in that wallet could be spent or transferred.

Why it was flagged

A raw wallet private key grants broad signing authority over the wallet, not just narrowly scoped marketplace access.

Skill content
export CHEESE_PRIVATE_KEY="0x..."  # Your wallet private key
Recommendation

Use a dedicated low-balance wallet for this skill, avoid exposing a primary wallet private key, and confirm every transaction before signing.

ConcernMedium Confidence
ASI02: Tool Misuse and Exploitation
What this means

The agent could lock, release, or otherwise move real funds if used without careful supervision.

Why it was flagged

The documented workflows authorize financially significant blockchain actions, but the provided instructions do not define explicit spend caps, confirmation requirements, or transaction-approval boundaries.

Skill content
Create request — Post job with ETH/USDC escrow + required collateral ... Accept request — Deposit required collateral ... Complete — Release escrow to provider
Recommendation

Require explicit user approval for each on-chain transaction, set maximum spend/collateral limits, and review request addresses and amounts before signing.

What this means

Users cannot verify from the supplied skill artifacts what code will handle the private key and transactions.

Why it was flagged

The skill relies on executing a local TypeScript CLI, while the supplied artifact set contains no code files or install spec to review that implementation.

Skill content
A unified CLI is available at `~/clawd/cheese/scripts/cheese-cli.ts`: ... `npx tsx scripts/cheese-cli.ts <command> [options]`
Recommendation

Install the CLI only from a trusted source, inspect it before use, pin a known commit/version, and avoid running it with a wallet holding significant funds.

What this means

Sensitive task details could be shared with a counterparty or over the marketplace communication channel if the agent includes them in chat.

Why it was flagged

The marketplace intentionally depends on Waku messages for coordination with counterparties, which may include work details, delivery confirmations, or dispute-related information.

Skill content
YOU MUST USE WAKU CHAT FOR ALL REQUEST COMMUNICATION.
Recommendation

Do not send secrets or unrelated private data in Waku chat, verify the request address, and treat counterparty messages as untrusted instructions.

NoteHigh Confidence
ASI10: Rogue Agents
What this means

A monitoring process may continue running and reacting to marketplace communications until the trade ends.

Why it was flagged

The skill explicitly asks for a long-running watcher after creating or accepting a request. This is disclosed and purpose-aligned, but users should notice the ongoing activity.

Skill content
Immediately run: `npx tsx scripts/cheese-cli.ts chat read <request_address> --watch` ... Keep monitoring until the request is completed or cancelled
Recommendation

Start the watcher only for intended requests and stop it after completion, cancellation, or expiry.