Seatbelt for Governance
Seatbelt automates proposal simulations and translates technical details into clear, actionable insights for all governance participants.
What is Seatbelt?
Seatbelt is an automated simulation layer that ScopeLift and Uniswap Labs built to pre-screen every on-chain proposal before execution. The tool replays a proposal in a local fork, captures all state changes and external calls, and outputs a PDF or HTML report in plain language. All processing occurs off-chain, so the simulation adds no gas cost and does not touch mainnet.
A small error in a proposal’s calldata like an incorrect parameter and/or a misplaced address can brick a treasury or block future upgrades. Seatbelt provides a final, automated check. In past Compound votes (examples 1, 2, and 3), the simulator identified transactions that would have reverted on execution, allowing proposers to fix the issue before the vote closed.
How seatbelt works
Every few hours Seatbelt scans supported DAOs for active proposals, forks the chain at the current block, and runs each candidate through Tenderly. The resulting trace is parsed for:
State-variable mutations
External contract calls and target addresses
Emitted events
Compiler warnings
Static-analysis findings from Slither (by Trail of Bits)
The fork can override timestamps, block numbers, and quorum thresholds, so Seatbelt handles both Governor Bravo and OpenZeppelin Governor instances without manual configuration.
Stakeholder impact
Delegates and token-holders receive a one-page summary that describes the intended changes in ordinary language, removing the need to inspect raw calldata.
Proposal authors get a low-overhead linting step to catch mistakes before submission.
Governance and operations teams see fewer failed executions and less emergency remediation work.
Seatbelt implementation
Seatbelt currently runs on Uniswap, Compound, and ENS. Additional DAOs can opt in by submitting a pull request to the Seatbelt repository or by contacting Tally. The same simulation routine can be executed locally for ad-hoc checks; the local output matches the report served to delegates.
Helpful resources
Last updated
Was this helpful?