harness-web3

What problem does it solve?

It prevents Web3 AI systems from making risky or incorrect decisions when data is missing, evidence is unclear, rate limits are exceeded, or transaction-writing actions could be irreversible.

Core Features & Use Cases

• UNKNOWN correctness discipline: Forces on-chain and external-data signals to return UNKNOWN (never coerced to 0/null/neutral) and to remain visible, preserving scoring integrity in production.
• Rate limit budget enforcement: Ties every external call to the declared budget in docs/web3/chain-config.md, requiring updates before adding new API usage.
• Evidence-required risk scoring: Requires each risk/score to cite specific on-chain evidence sources (RPC/BaseScan endpoints) and document availability assumptions in ADRs.
• Irreversibility gates for Tier 5 actions: Blocks automatic execution of deployment, signing/submission to mainnet, wallet/key changes, and ABI changes, requiring explicit human authorization and second-human sign-off patterns.

Quick Start

Install harness-governance and harness-web3 in your project, then use the skill’s rules to ensure any Web3 scoring pipeline returns UNKNOWN when data cannot be computed and never auto-executes Tier 5 transaction-signing workflows.