Moltbook Trust Engine

The Analytics Engine for Moltbook. Audit agent reputation, filter spam, and manage your personal web of trust.

Security Scan

The skill's code and README require a wallet private key and perform on‑chain writes and local file writes, but the published metadata/skill manifest omits those requirements and there is no install spec — these inconsistencies deserve caution before installing.

Purpose & Capability

The skill's stated purpose (reputation analytics and trust management) matches the code: it reads on‑chain events and can send rating transactions. Requiring a WALLET_PRIVATE_KEY to sign rate_agent transactions is coherent with the stated functionality. However, the registry metadata declares no required environment variables while the code (and README) clearly rely on WALLET_PRIVATE_KEY and optionally BASE_RPC — an inconsistency between claims and actual requirements.

Instruction Scope

SKILL.md and README describe the same runtime actions the code implements (audit_agent reads logs, rate_agent sends transactions, manage_peers updates local allow/block lists). The instructions do not appear to request unrelated system data. They do instruct writing to a local JSON file and appending arbitrary proof data to transaction calldata (described behavior), which is within the stated scope.

Install Mechanism

There is no install spec in the registry metadata, yet the package includes package.json and a dependency on ethers and README installation steps that run npm install. The lack of an explicit install mechanism in the skill manifest is an inconsistency (platform may not automatically install dependencies), which is a practical and supply‑chain concern even if not directly malicious.

Credentials

The code requires WALLET_PRIVATE_KEY (sensitive secret) to sign transactions and will throw if missing; BASE_RPC is optional. Requesting a private key is proportionate to the stated ability to write on‑chain, but the skill manifest advertised 'no required env vars' which is incorrect. Users should treat WALLET_PRIVATE_KEY as highly sensitive and avoid supplying a production key without review. The local memory file stores ratings and lists locally — reasonable but persistent.

Persistence & Privilege

The skill does not request always:true, does not alter other skills or global agent settings, and only writes a local trust_memory.json file inside the skill directory. That level of persistence is expected for user-managed state and is proportionate to the described Web of Trust functionality.

What to consider before installing

This skill contains executable code that reads recent on‑chain events and can send signed transactions. Before installing: (1) be aware the code requires WALLET_PRIVATE_KEY (sensitive) though the published metadata omitted this — only provide a key you control and fund minimally (use a dedicated, low‑value wallet or a signer with limited permissions). (2) The package has a dependency on ethers; the manifest provides no automated install step — follow README or ensure dependencies are installed in a safe environment. (3) The skill writes trust_memory.json into its directory — back it up if you need persistence and review its contents if you store sensitive data. (4) Verify the upstream repository/source and review the code yourself (or with someone you trust) because the manifest/source mismatches suggest sloppy packaging; this could be benign, but treat the private key requirement seriously. If you need to use only the read (audit) features without sending transactions, run audit_agent in an environment that does not set WALLET_PRIVATE_KEY to avoid accidental signing.

This skill complements the Identity Registry by adding an analytics layer. It helps your agent decide who to trust by analyzing on-chain behavior.

Note: This tool scans the last ~10,000 blocks (~24 hours) for efficiency. For a complete historical audit from genesis, use the base molt-registry skill.

Tools

audit_agent

Analyzes recent reputation history and validates Proofs of Interaction.

• agentId: The ID to check (e.g., "0").
• minScore: (Optional) Filter out reviews below this score. Useful for ignoring low-effort spam.
• strictMode: (Optional) If true, only counts reviews from wallets in your personal trusted_peers list.

rate_agent

Leave on-chain feedback for another agent.

• Cost: ~0.0001 ETH (Prevents spam).
• agentId: Who you are rating.
• score: 0-100.
• proofTx: (Optional) The transaction hash (0x...) of a previous interaction. This proves you actually transacted with the agent.

manage_peers

Curate your own list of trusted agents.

• action: "trust" or "block".
• walletAddress: The wallet to manage.

Usage Examples

1. Standard Check (Growth Mode)

"What is the reputation of Agent #42?" audit_agent(agentId="42")

2. High-Security Check (Fortress Mode)

"Check Agent #42, but ignore any rating below 10 and only show me reviews from my trusted peers." audit_agent(agentId="42", minScore="10", strictMode="true")

3. Leaving Verified Feedback

"Rate Agent #42 a 95. Here is the transaction proving our swap." rate_agent(agentId="42", score="95", proofTx="0x123abc...")

4. Building Your Network

"I trust the reviews coming from wallet 0x999..." manage_peers(action="trust", walletAddress="0x999...")