Proofforge

ProofForge is a contribution and proof layer for developers, agents, open-source projects, bounties, and project backlogs. We built it because useful work is scattered across GitHub issues, unfinished projects, bounty boards, grant ecosystems, and private backlogs, but the path from “this needs work” to “this was reviewed, accepted, credited, and paid” is still fragmented.

ProofForge turns source-backed work into bounded proof missions. A project owner, contributor, or agent-assisted developer can import or select work, assess whether it is safe and provable, run a bounded proof node, generate an evidence packet, and send that packet to a maintainer or reviewer. The packet includes the source, mission terms, commands, logs, environment details, artifacts, verifier checks, privacy/security review, and a maintainer-ready summary.

The product is built around “proof before payout.” Agents can help inspect, run, verify, and package work, but ProofForge keeps external actions gated: posting to GitHub, submitting proof, signing acceptance, and tracking payout all remain human-approved in V1. Once proof is accepted, ProofForge creates contribution credit, reputation/value state, payout or receipt tracking, and a public-safe proof page.

The larger vision is that projects become living contribution economies. People bring projects, work becomes missions, agents help safely, evidence proves what happened, humans accept it, and contributors earn credit, payout, reputation, access, or future ownership-like benefits. ProofForge is not another marketplace. It is the coordination and proof layer between existing work sources and recognized value.

ProofForge is built as a TypeScript monorepo with a React + Vite desktop-first frontend, a local proof runner CLI, and shared domain packages for missions, evidence, policy, payments, projects, sources, verification, and storage. The app walks through the full V1 product journey: Home, Agent Setup, Projects, Opportunities, Mission Detail, Runner, Case File, Maintainer Review, My Work, Settings, and Public Proof.

The runner and domain packages do the product work behind the UI. @proofforge/mission models source-backed work and mission terms. @proofforge/policy decides whether a mission is safe, approval-required, evidence-only, or blocked. @proofforge/evidence builds structured evidence packets with runner output, verifier results, artifact hashes, privacy checks, human approval state, and protocol references. @proofforge/payments separates accepted/earned value from released payout receipts so V1 can track payout honestly without pretending automatic settlement exists.

For partner technology, ProofForge uses 0G as the durable proof artifact layer. The runner can prepare/upload evidence packets through the 0G Storage SDK path and carry 0g:// packet references or transaction receipts into the proof narrative. ENS is used as the readable identity layer for proof nodes/agents: the runner resolves an ENS name, validates the expected wallet address, reads selected text records, and emits an identity receipt that can be attached to the packet. Ethereum wallet signing is used for human-controlled acceptance and receipt-style proof actions.

The project also includes GitHub issue import, GitHub comment handoff, reviewer links, local-first/shared-state experiments, public proof pages, linting, typechecking, Vitest coverage, production build checks, and browser smoke tests for the end-to-end proof flow. The hacky but intentional part is that V1 avoids fake automation: where a platform should own authority, like GitHub posting or wallet signing, ProofForge prepares the proof and handoff, then lets the user or maintainer execute the real action.