Programmable v4 Builder v0.4.0
The first standalone release of Programmable v4 Builder: a portable Agent Skill that turns a plain-language product idea or existing repository into a checked, reviewable Uniswap v4 project and exact GitHub application.
Install
gh skill install 0xprogrammable/programmable-v4-builderPinned installation:
gh skill install 0xprogrammable/programmable-v4-builder programmable-v4-hook-builder@v0.4.0 --agent codex --scope userWhat ships
- Four open starters and 33 composable capability packs for ordinary launches, custom hooks, games, maps, wallet quests, SDK clients, liquidity systems, external services, novel architectures, and more.
- A progressive knowledge router that expands public Lego pack ids and loads only the relevant Uniswap, Programmable, security, runtime, SDK, or workflow context.
- Pinned official-source records across v4 core, periphery, SDKs, Universal Router, Permit2, OpenZeppelin hooks, Uniswap AI, and deployment references.
- Closed submission schemas, semantic checks, fee-conformance tooling, source binding, GitHub application preparation, read-only status, and explicit two-step external actions.
- The inclusive Programmable 10 bps policy, with immutable claim authority
0x4957f49620AFf3Adbbe8195a4f633E49cc93376cseparated from platform administration.
Release evidence
- Public CI passed on Node 20 and 22.
- 34 Solidity unit, fuzz, and invariant tests passed.
- Slither 0.11.5 ran 101 detectors with zero remaining findings; CodeQL passed.
- The exact public tag was installed and verified in fresh Codex, Claude Code, and GitHub Copilot targets.
- The official-source drift check reported no findings across 23 repositories, 2 feeds, 32 deployment records, and 16 source artifacts on 2026-08-02.
Release assets contain a deterministic skill archive, file-level SHA-256 manifest, SPDX 2.3 SBOM, release-state receipt, and SHA256SUMS.
Honest boundary
This release is a Builder and review system. It is not an independent security audit, deployment, provider approval, Uniswap endorsement, or guarantee that generated projects are safe. Model-backed prompt evaluations were not run because they require an explicit provider credential and may cost money. Every generated project still needs exact review and evidence for its own source, configuration, dependencies, and runtime.