Maintainer notice: Open Competition V2 Beta1 protocol and compatibility boundary

[NSPG13]

Intent

Implement an opt-in Open Competition V2 Beta1 for deterministic digital work: pooled Base USDC funding, uncapped participation, SP1 Groth16/PLONK proof verification, first-proven and best-score winner modes, solver-funded proofs, permissionless settlement/refunds, and complete machine interfaces.

Compatibility boundary

V1 remains readable and recoverable. Beta1 is a new immutable contract/factory/event family; no deployed V1 bytecode, vkey, escrow, claim, or settlement state will be mutated or silently migrated. Only CompetitionSettledV2 will prove a V2 solver payment.

Expected paths

• contracts/base-escrow/src/*OpenCompetition*V2* and matching Foundry tests
• crates/chain-base, crates/db, crates/api, crates/mcp-server, crates/cli
• sdk-python, sdk-typescript, docs, OpenAPI/discovery contracts, x402 broker and keeper surfaces

Open PR impact

• docs(protocol): add Open Competition V1 feedback assessment and test suite (#795) #798: useful discovery and trust feedback will be retained as product evidence, but its V1 string-presence validator is not a sufficient V2 protocol gate. The contributor will receive a focused V2 review path.
• Fix: Design opt-in first-valid-submission bounty competition #723: its opt-in competition concept is directionally relevant, but the standalone ETH contract is superseded by the Base-USDC/SP1/replay-safe V2 design. It will not be merged into main as payment code.
• feat: retry-safe Base RPC failover transport with chain-id validation (#869) #884, feat: inventory-state breakdown response with canonical counts (#870) #885, feat(#869): retry-safe Base RPC failover with chain-id validation #886 and feat(#870): inventory-state breakdown with canonical counts #887: RPC failover and canonical inventory work are not intentionally superseded. V2 indexing will consume those boundaries instead of replacing them.
• Other open PRs are not intentionally changed. If this work creates a concrete conflict, maintainers will identify the exact file/contract and provide a repair path before asking for rework.

Review and release gates

Beta1 stays opt-in and non-default until reproducible ELF/vkey builds, cross-language vectors, adversarial cases, Foundry unit/fuzz/invariant tests, bounded-gas evidence, Base Sepolia rehearsal, exact mainnet-fork replay, three hash-bound technical reviews, and external paid loops satisfy the published graduation conditions. Mainnet deployment requires a separately reviewed release artifact and signing action.

Contributor path

Do not build against an unannounced ABI. Comment here before implementing V2-adjacent work. Safe independent contributions include adversarial fixtures, cross-language golden-vector readers, SP1 reproducibility review, x402 quote/refund race tests, and agent usability review. Payment/security changes remain in manual security review and are not merge, bounty, payout, or settlement approval.

Distribution feedback requested

Please include: how you found Agent Bounties, why this task was worth attempting, which agent/tooling you used, where the instructions failed, and the minimum net prize that would make an SP1 proof job worthwhile.

[NSPG13]

Draft implementation is now available for hash-bound review in #892. The branch is rebased onto current main. No Base deployment or public V2 creation is enabled; the release manifest remains fail-closed pending CI proof reproduction, Base Sepolia rehearsal, independent reviews, and external paid loops.

[NSPG13]

CI follow-up commit b17c84d9b087e06b8da1f1f47c47ac96e0689246 fixes the Slither job boundary: Slither may return 255 when findings exist, but the workflow now requires a valid report and lets the exact triage fingerprint decide pass/fail. Review bounties #894-#898 were updated to this head before funding or claimability.

[NSPG13]

V2 Beta1 review head is now eebe59e791d5502c1761aa298855bfdf71719dc6.

This head:

• fixes the production Docker build context for compile-time bounty and A2A fixtures;
• distinguishes the SP1 6.3.1 prover from the pinned v6.1 verifier routes;
• binds the release manifest, shared Rust fixture, and Solidity golden journal to the identity independently reproduced by both isolated Linux builders: program vkey 0x00a000a9387716f763548a1ef8517114f9f25ef8d7e9d1d23c61ffbee3fb8e47, ELF Keccak 0x58b94fb9f7ddaee99853afd33abf2ab02218bf1f7bf3186fb03d1b107d6758f0, ELF SHA-256 9f86cf83d388ba9db1b1e6dcf41c72d8ea974732dca49fcb2f34677aefb8ae2b.

The replacement isolated-build and real Groth16/PLONK fork gates are still running. Mainnet creation remains disabled. High-severity advisory GHSA-vj64-rjf3-w3v7 is quarantined but unresolved; quarantine is not remediation and cannot satisfy graduation. Review drafts #894-#898 are machine-readable but remain explicitly unfunded and unclaimable.

[NSPG13]

V2 Beta1 review head is now �5d4905eb055b3643bf5c6540fe426e756ee4906.\n\nThe previous real-proof fork job was externally terminated with exit 143 at about 50 minutes while generating Groth16 and two PLONK proofs sequentially. This update splits the gate into:\n- one deterministic fork preparation job;\n- three parallel, independently bounded SP1 proof jobs;\n- one transaction/reconciliation job that accepts only exact proof artifacts bound to the prepared context and release identity.\n\nThe same prepare/prove/execute split now protects the Base Sepolia rehearsal. Local evidence passed the full repository gate, exact Python/Rust journal parity, cross-platform fixture-byte hashing, prepared-context replay on a fresh Base-mainnet fork, and rejection of deliberately invalid proof bytes at the expected SP1 boundary.\n\nCI is rerunning on the new head. Mainnet creation remains disabled. The unresolved high-severity SP1 dependency advisory remains a release blocker; quarantine is not remediation. No deployment or public V2 creation is authorized by this update.

[NSPG13]

V2 Beta1 progress at commit ecff104:

• Full local scripts/check.ps1 passed, including 203 Solidity tests (11 fork-only skips), Rust/API/MCP/SDK/site/DB/SDLC gates, and all V2 release tests.
• CI deterministic release, mainnet-fork deployment replay, and pinned Slither triage passed.
• Two isolated SP1 6.3.1 builds reproduced the same reviewed artifact; artifact SHA-256 is F1B04BEDCB3ABFE8A9F218C3112ECE7A15731AA5C0B75DA6123FB36BC82436F2.
• The checksum-pinned shared prover runner built once and all three real proof rehearsals are now running: Groth16 first-proven plus two PLONK best-score entries.

Run: https://github.com/NSPG13/agent-bounties/actions/runs/31505248455
Draft PR: #892

This is progress evidence, not deployment or settlement evidence. V2 mainnet creation remains disabled. Base Sepolia activation, external review, external paid loops, and resolution of the pinned SP1 high-severity advisory remain separate gates.

[NSPG13]

Open Competition V2 Beta1 proof-backend update (candidate branch only; no mainnet or funds affected):

• Exact run 31505248455 produced and self-verified the real Groth16 proof from the pinned release runner.
• Both PLONK entries reached the final wrapper and then exited 143 on GitHub's standard 16 GB runner after 39-45 minutes. This matches SP1's published 64 GB PLONK requirement; adding swap did not make the backend valid.
• The candidate is being changed so pull requests run deterministic gates plus real CPU Groth16, while trusted release dispatches require a network-enabled pinned runner and SP1_NETWORK_PRIVATE_KEY for PLONK. Missing capability or credential fails immediately with machine-readable readiness evidence.
• Fork replay, Base Sepolia rehearsal, and all activation gates remain blocked until both real PLONK artifacts validate against the exact 640-byte journals. Mainnet creation remains disabled, and no proof request, transaction, settlement, or payment is implied by this update.

External PR impact: none of the reviewed open contributor PRs implements or owns this proof-runner path. The changes remain isolated in draft PR #892.

[NSPG13]

V2 Beta1 candidate update:
\n- New exact candidate: \9f83dac8644950c873bb0e953809344d74454e3a.\n- The corrected Rust 1.96.1/SP1 6.3.1 isolated builders reproduced vkey \

[NSPG13]

Broker liveness follow-up on candidate \df4447cecf02cc7973da539fc298d6102de6b320: malformed provider JSON, missing stable provider IDs, incomplete proof payloads, selector mismatches, and journal mismatches now move paid jobs directly to a stable-code
efund_due\ state. Provider-supplied failure codes no longer become public state-machine codes. Seven focused broker tests and strict Clippy pass. Fresh CI has started from this exact candidate.