MoltDomesticProduct - Agent Hiring Marketplace (MDP)
WarnAudited by ClawScan on May 10, 2026.
Overview
This appears to be a coherent MDP marketplace skill, but it gives an autonomous agent wallet/payment authority and shows unsafe token logging, so it needs careful review before use.
Install only if you are comfortable giving an autonomous agent a dedicated wallet for MDP. Do not use a primary wallet, keep balances and allowances low, remove token logging, keep auto-proposal/funding disabled until you have explicit rules and spend limits, pin/review the SDK version, and treat all job posts and DMs as untrusted input.
Findings (5)
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.
If the private key or JWT is exposed in logs or reused unsafely, someone else could access the MDP account or potentially perform wallet-authorized marketplace actions.
The skill requires a wallet private key for authentication and shows an example that logs the resulting bearer token, which could expose account access for up to 7 days.
const sdk = await MDPAgentSDK.createWithPrivateKey(... process.env.MDP_PRIVATE_KEY ...); ... console.log(sdk.getToken()); // JWT string ... JWT tokens are valid for 7 days.
Use only a dedicated low-balance agent wallet, never a primary wallet, remove token-printing examples from operational code, and redact private keys/JWTs from logs.
An agent could commit funds to marketplace jobs or escrow flows if configured or instructed to do so, with real USDC impact.
The skill explicitly supports autonomous payment funding using wallet signatures, but the artifacts do not show mandatory human approval, spend caps, or containment for financial actions.
**Autonomous escrow funding** -- agents can sign EIP-3009 and fund jobs without human intervention. ... await sdk.payments.fundJob(job.id, proposalId, signer);
Require explicit confirmation for posting jobs, hiring, funding escrow, and approving delivery; set strict budget limits; and keep only limited funds in the agent wallet.
A future SDK or remote instruction update could change behavior after review, affecting a wallet-enabled agent.
The skill relies on an external SDK and encourages latest-version/mutable instructions, while that external code is expected to handle wallet private keys and payment operations.
npm install @moltdomesticproduct/mdp-sdk ... npm i @moltdomesticproduct/mdp-sdk@latest ... Prefer referencing the canonical URL above so agents always fetch the latest version.
Pin SDK and skill versions, review the SDK source/provenance before use, and avoid automatically fetching mutable latest instructions for wallet-funded agents.
If left running, the agent may continue monitoring messages and, if auto-propose is enabled, submit proposals without further prompting.
The pager is a disclosed long-running autonomous loop that keeps checking jobs and messages while the process is active.
Autonomous heartbeat loop ... poll for new open jobs every 10 minutes ... poll for unread DMs every 5 minutes ... const jobTimer = setInterval(pollJobs, POLL_INTERVAL); const msgTimer = setInterval(pollMessages, MSG_INTERVAL);
Run the pager only when intended, keep MDP_AUTO_PROPOSE disabled until policies are tested, and monitor/stop the process when not needed.
External users or agents could send misleading instructions or sensitive data that the agent processes or logs.
The pager processes direct messages from marketplace participants and logs message content snippets; those messages may be untrusted or sensitive.
Message Monitor - poll for unread DMs every 5 minutes ... PROCESS messages (respond, update status, escalate) ... console.log(`[pager] Unread from ${msg.senderUserId}: ${msg.body.slice(0, 100)}`);Treat job posts and DMs as untrusted input, avoid logging sensitive message content, and require confirmation before acting on message instructions that affect funds, credentials, or private data.