FlexGov

FlexGov is a governance-observability and decision-support system for DAOs. Governance platforms can show that a proposal passed under their configured rules, but they often do not clearly show how concentrated the participating voting power was, which wallets shaped the result, or whether the outcome depends on the weighting mechanism. FlexGov converts proposal and ballot data into a canonical Governance Health Report. Its deterministic TypeScript engine measures observable signals including largest-wallet share, top-three concentration, Gini inequality, the minimum number of wallets controlling half of cast voting power, timing patterns and—when authoritative data is available—quorum progress. Missing evidence such as eligible-member turnout, delegation history, wallet relationships and token-acquisition history is explicitly marked unavailable rather than estimated. The engine then recounts the same observed ballots under four conditional weighting assumptions: one-token-one-vote, square-root/quadratic weighting, one-wallet-one-vote and an experimental Gini-adaptive transformation. FlexGov classifies whether the recorded winner is robust, contested or rule-dependent across these tests. These are transparent stress tests, not predictions of how participants would behave under different binding rules and not proposed retroactive replacements for a DAO’s constitution. FlexGov’s reference on-chain integration uses The Graph’s decentralized Query API to retrieve Compound Governor Bravo proposal 393 on Ethereum. Its reusable server-side CLI cursor-paginates and validates all 29 indexed ballots, then captures authoritative quorum, queue and execution lifecycle, contract actions, subgraph deployment and indexed-block provenance. The resulting report shows 723,624.67 COMP of voting weight against a 400,000 COMP quorum, a 30.8% largest-wallet share, a 52.9% top-three share and a robust “For” outcome under all four tested transformations. Every report separates source facts, deterministic findings, data availability, limitations, advisory interpretations and verification metadata. Reports are canonically serialized and assigned SHA-256 source, configuration and report hashes. These hashes establish integrity and reproducibility boundaries; they do not prove that an interpretation is politically correct or that unusual activity demonstrates wrongdoing. The public prototype also supports current Snapshot proposal exploration and pinned reports for ENS, Arbitrum, Uniswap, Gitcoin, Sushi and Optimism. FlexGov does not claim that concentration or timing correlations prove bribery, collusion, Sybil control or malicious intent. It is advisory by default and does not silently change a DAO’s rules or enforce responses. FlexGov additionally ships a public agent SKILL and reusable Graph-backed CLI. An external agent such as Codex can follow the SKILL, retrieve live Compound governance data, generate the deterministic canonical report and explain it under four evidence-bounded headings: Graph-sourced facts, deterministic findings, unavailable data and limitations, and clearly labelled advisory interpretation. The LLM does not calculate the metrics, generate a hidden risk score or decide whether a proposal should pass. The wider adaptive-safeguards vision is for communities to pre-authorize policies such as warnings, challenge periods, execution delays or human review when declared conditions are met. Those enforcement mechanisms remain future architecture; the submitted MVP supplies transparent evidence, counterfactual analysis and reproducible reports. Pre-existing research disclosure: FlexGov was informed by governance research and conceptual work I completed during my MSc Data Science before the event, including an academic paper exploring related governance problems. No pre-existing FlexGov software implementation, interface, design assets, analysis engine, report schema, source adapter, Graph integration or agent tooling was submitted. All submitted product code and functionality—including the TypeScript engine, canonical reports and hashing, Snapshot interface, Compound/The Graph integration, agent SKILL and CLI, tests and deployment—were created during the hackathon. The whitepaper was drafted during the event from that earlier research and combines the implemented MVP with a clearly labelled wider vision.

FlexGov consists of a zero-runtime-dependency TypeScript governance-analysis engine, source adapters, a versioned Governance Health Report schema, a reusable command-line tool and a React/Next.js reference interface deployed through GitHub Pages. Governance sources are normalized into a common proposal context and ordered ballot-event format. The engine measures voting-power concentration, Gini inequality, controlling-wallet count, timing patterns and authoritative quorum where available. It then recounts the same ballots under one-token-one-vote, square-root/quadratic weighting, one-wallet-one-vote and experimental Gini-adaptive weighting, classifying whether the recorded winner is robust, contested or rule-dependent. The reference on-chain integration uses The Graph’s decentralized Query API to retrieve Compound Governor Bravo proposal data from Ethereum. A secret-safe server-side CLI cursor-paginates every indexed ballot, verifies that the retrieved count matches the proposal vote total and fails instead of generating a partial report. It also captures quorum, proposal lifecycle, contract actions, subgraph deployment, indexed block and indexing-error status. The Graph API key remains server-side and is never included in browser code, generated reports or logs. The resulting Governance Health Report separates source facts, deterministic findings, data availability, limitations, advisory actions and verification metadata. Objects are recursively serialized with stable key ordering and hashed using SHA-256 through the Web Crypto API, producing source-data, configuration and report hashes. The public interface consumes a precomputed canonical Compound report generated from live Graph data; the CLI can independently generate fresh JSON without using the website. FlexGov also includes a public agent SKILL.md describing how external coding agents can invoke the Graph-backed CLI and consume its output. In the demonstrated AI workflow, an agent runs the tool and explains the resulting report under evidence-bounded headings, while all metrics and classifications remain deterministic rather than LLM-generated. Snapshot Hub provides the second adapter path, supporting current proposal exploration and pinned reports across six DAOs. The repository includes 34 engine tests plus dedicated tooling tests for cursor pagination, completeness failures, provenance, canonical hashes and secret safety.