# Genuine parity acceptance matrix

User authorization: continue until genuine ETHSkills parity, with recipes for Pharaoh, Aave V4, LFJ, Blackhole, Silo, Euler and BENQI. Started 2026-09-06. This file records acceptance criteria and evidence, not a claim of completion.

Reference inventory: `research/ethskills-reference/inventory.json` pins all 21 advertised topic files by URL, byte size and SHA-256. Parity means equivalent useful capabilities for Avalanche, not copying Ethereum-specific recommendations, word counts or checklist counts. Public distribution is a separately authorized action; its absence remains explicit.

## Gates

| Capability in ETHSkills | Required Avalanche result | Acceptance evidence | Status |
| --- | --- | --- | --- |
| Why / chain selection / L2s | Honest C-Chain versus sovereign L1 decision, operational/security tradeoffs | Decision examples plus independent task evaluation | Complete within scope |
| Ship / orchestration / frontend playbook | Complete local contract-to-UI-to-release path, coherent network configuration and recovery | Runnable example contract, frontend state machine, local deployment and browser flow | Complete within scope |
| CROPS architecture review | Custody, censorship, source/license, privacy, admin and exit review suitable for C-Chain/L1s | Worked architecture review and evidence template | Complete within scope |
| Protocol / gas / tools | Dated ACP activation, actual RPC/compiler behavior and executable fee checks | Pinned network/VM/tool versions; real reads and local tests | Complete within scope |
| Wallets / standards / concepts | ERC20/721/1155/4626, signed authorizations, app incentives, automation and Avalanche compatibility | Compiled examples, authorization/math/failure tests, supported-feature lookup | Complete within scope |
| Building blocks / addresses | Seven requested protocol-specific recipes and current deployment registry | Primary source commit, chain ID, block hash, code hash and semantic-call results | Complete within scope |
| Security / audit | Domain-organized threat review with demonstrated exploit/invariant evidence | Independent read-only contract review, compatibility/conservation regressions, scoped findings report | Complete within scope |
| Noir/privacy | Working privacy example, public-input/nullifier/recipient binding, current toolchain and integration limits | Compiled/executed circuit and proof/verification path where supported | Complete within scope |
| Testing / QA | Unit/fuzz/fork/negative-path procedures that an agent can execute | Actual test execution, not instructions alone | Complete within scope |
| Indexing | Resumable event ingestion, deduplication, provenance and unavailable/stale states | Executable ingestion/replay/failure tests | Complete within scope |
| Human and agent distribution | Discoverable routing, focused skills and bundled references/examples | All links/hashes/package assets verified; browser tests with expanded library | Complete within scope |
| Model-gap methodology | Independent agents produce artifacts under objective acceptance tests | Preregistered tasks, baseline/skill-assisted results, failures repaired and rerun | Complete within scope |

## Seven protocol recipe gates

Each requested protocol must include a pinned ABI/version, primary sources, an Avalanche deployment record, live read-only semantic evidence, an unsigned operation planner, meaningful invalid-input/role/version/limit tests, and at least one exercised integration path. Signatures, public transactions and production changes are outside the current authorization; local-fork writes use isolated unlocked test accounts only.

| Protocol | Required distinction | Status |
| --- | --- | --- |
| Pharaoh | Current CL/DLMM and x(3,3) contracts versus historical deployments | Local-executed |
| Aave V4 | V4 Hub/Spoke/asset IDs; no V3 ABI substitution | Local-executed |
| LFJ | Liquidity Book router/path/bin version | Local-executed |
| Blackhole | AMM/CL distinction, gauges and veNFT ownership/approval model | Local-executed |
| Silo | Exact deployment implementation and per-market collateral/debt behavior | Local-executed |
| Euler | EVK/EVC and per-vault/deprecation/collateral-controller state | Local-executed |
| BENQI | Lending markets versus liquid staking; ERC20 versus native AVAX entry points | Local-executed |

## Evidence labels

- **Documentation-sourced:** a primary source names a deployment or behavior.
- **Runtime-checked:** bytecode and relevant semantic calls were observed at a recorded chain/block.
- **Local-executed:** an operation or failure path was executed on an isolated local node or fork; not a mainnet transaction.
- **Agent-evaluated:** a separate agent produced an artifact that passed an independent executable grader.
- **Blocked/unverified:** the required proof is unavailable; it cannot count toward a passing gate.

Website test totals are never substituted for contract, cross-chain, privacy or agent-performance evidence. Passing a representative evaluation is not a universal accuracy guarantee or a third-party security audit.

## Acceptance results — 2026-09-06

- **Breadth:** root +30 focused topics cover the21 ETHSkills reference topic capabilities, with Avalanche L1/ICM-specific guidance and seven protocol procedures. Functional scope, not word/checklist counts, defines this mapping.
- **Protocols:** all seven have pinned manifest/ABI provenance, observed bytecode/semantic bindings, unsigned planning, negative tests and successful representative isolated-fork operations.76 protocol tests pass.
- **Core:**15 Solidity tests,512 fuzz runs (including256 three-user48-action sequences),14 Node tests,15 real browser/local-EVM/indexer checks, zero page errors.
- **Standards:**26 tests plus live USDC domain/nonce, ERC8004 linkage and vault observations. No signature or real x402 settlement is claimed.
- **Privacy:**5 circuit tests, actual proof generation and independent backend verification, generated Solidity verifier and stateful membership/nullifier checks on local EVM.
- **Agent evaluation:** fresh baseline and skill-assisted implementations each pass the same24 fixed LFJ checks on first grading. One task/sample per arm; no measured uplift, general model superiority or direct ETHSkills benchmark claim.
- **Integrated verification:** all12 commands in research/parity-verification/summary.json passed, including all seven protocol forks and the real core/Noir flows.
- **Distribution:**31 guides, code, manifests, tool hashes and dated evidence are packaged; final clean-extraction and website QA results are recorded in qa-artifacts and RELEASE.md.

### Review-method deviation, recorded explicitly

The optional seeded-exploit exercise was blocked by automatic cyber review and stopped. It is **not** counted as an executed or passing exploit benchmark, and its unexecuted fixtures are excluded from distribution. The security gate instead uses a narrower independent read-only review of the real reference contract, followed by ordinary compatibility/conservation regressions. That review identified a rebasing documentation overclaim and missing test coverage; both were addressed. The original prerequisite was revised to this safer evidence method, not silently marked passed. See research/parity-audit/read-only-review.json and review-resolution.json.

### Boundaries that remain outside this release

This is functional guide/reference-example parity, not exhaustive ecosystem or operation coverage. Successful funded Silo/Euler borrowing, mature sAVAX redemption, governance/claims/LPs, real wallet extensions, full L1 validator/ICM/ICTT execution, actual paid HTTP settlement and public deployment remain unverified. The corresponding guides explain the required procedure and explicitly distinguish it from demonstrated execution. No public deployment, signing, repository publication or third-party audit has occurred.