{"id":50795316,"url":"https://github.com/dpdanpittman/tribunal","last_synced_at":"2026-06-12T14:01:33.883Z","repository":{"id":357516170,"uuid":"1237182949","full_name":"dpdanpittman/tribunal","owner":"dpdanpittman","description":"Adversarial code-review methodology for LLM-assisted development, with on-chain reputation settled on Burnt XION.","archived":false,"fork":false,"pushed_at":"2026-05-22T18:24:57.000Z","size":2444,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-05-22T21:51:47.589Z","etag":null,"topics":["adversarial-review","ai-safety","burnt","code-review","cosmos","cosmwasm","golang","llm","methodology","reputation-system","smart-contracts","xion"],"latest_commit_sha":null,"homepage":"https://tribunal.mabus.ai","language":"Go","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"agpl-3.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/dpdanpittman.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-05-13T00:37:17.000Z","updated_at":"2026-05-22T18:25:01.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/dpdanpittman/tribunal","commit_stats":null,"previous_names":["dpdanpittman/tribunal"],"tags_count":4,"template":false,"template_full_name":null,"purl":"pkg:github/dpdanpittman/tribunal","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/dpdanpittman%2Ftribunal","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/dpdanpittman%2Ftribunal/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/dpdanpittman%2Ftribunal/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/dpdanpittman%2Ftribunal/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/dpdanpittman","download_url":"https://codeload.github.com/dpdanpittman/tribunal/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/dpdanpittman%2Ftribunal/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34247461,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-12T02:00:06.859Z","response_time":109,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["adversarial-review","ai-safety","burnt","code-review","cosmos","cosmwasm","golang","llm","methodology","reputation-system","smart-contracts","xion"],"created_at":"2026-06-12T14:01:32.827Z","updated_at":"2026-06-12T14:01:33.878Z","avatar_url":"https://github.com/dpdanpittman.png","language":"Go","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Tribunal\n\n\u003e _The unit of trust is not consensus — it is surviving adversarial scrutiny by identified agents whose history is on the public record._\n\nTribunal is a methodology and tooling for shipping LLM-assisted code without\ninheriting LLM failure modes. It composes three layers:\n\n1. A **process backbone** — state machine + spec-driven gates + role separation + git discipline.\n2. A **correctness toolkit** — adversarial multi-model review on top of cooperative-parallel lens review, plus a verification pyramid.\n3. An **on-chain incentive layer** — a soulbound reputation token on Burnt XION (CosmWasm) that tracks per-agent finding outcomes over time.\n\nThe methodology is a synthesis of two existing patterns:\n\n- [`reliqlabs/colosseum`](https://github.com/reliqlabs/colosseum) — adversarial spec review + Rust-flavored verification pyramid.\n- [`btspoony/mstar-harness`](https://github.com/btspoony/mstar-harness) — multi-role agent harness with spec-driven gates and lens-diverse QC trio.\n\nThe novel piece neither has: **reputation-weighted findings**. Agents have verifiable identities (ed25519 keypairs), every finding is signed and recorded, outcomes are settled by PMs/QA, and reputation deltas land on-chain.\n\n**Live**: [tribunal.mabus.ai](https://tribunal.mabus.ai) — methodology, [sample audits](https://tribunal.mabus.ai/audits/), and the [on-chain leaderboard](https://tribunal.mabus.ai/leaderboard) that queries the deployed contract on `xion-testnet-2` from the browser.\n\n## How it composes with clawpatch\n\nAs of v0.3.4+ ([ADR-0002](./docs/adr/0002-clawpatch-absorption.md), 2026-05-17), Tribunal's lens-parallel review stage can run through [clawpatch](https://github.com/openclaw/clawpatch) as a subprocess. The trust/discovery split:\n\n- **Clawpatch** owns discovery — heuristic + agent-based feature mapping, per-feature LLM review, fix-and-revalidate loops.\n- **Tribunal** owns trust — agent identity, ed25519-signed findings, adversarial multi-model review, PM/QA-settled outcomes, on-chain reputation.\n\nRun `tribunal review --via-clawpatch` and the lens stage dispatches via clawpatch instead of expecting skill-trio reports on disk. Findings come back through `internal/clawpatch/translate.go`, get signed by Tribunal-orchestrator, and land in the existing ledger. Two upstream PRs ([#64 `--prompt-file`](https://github.com/openclaw/clawpatch/pull/64), [#65 `--export-tribunal-ledger`](https://github.com/openclaw/clawpatch/pull/65)) added the integration hooks; both merged 2026-05-18. `tribunal fix --finding \u003cid\u003e` and `tribunal revalidate` round-trip state back to clawpatch via signed triage events.\n\n## Quick start\n\n\u003e Requires Go 1.23+. Optional: an Anthropic API key (`ANTHROPIC_API_KEY`) for the Claude adversary panel; v0.3+ adds Burnt XION on-chain settlement.\n\n```bash\ngo install github.com/dpdanpittman/tribunal/cmd/tribunal@latest\ntribunal init                            # scaffold .tribunal/ in the project\ncp tribunal.yaml.example tribunal.yaml   # tune panels + verify stack to taste\n\n# Verification pyramid (runs build / fmt / vet / test / ... per stack).\ntribunal verify .\n\n# Adversary review stage (lens-parallel trio is dispatched by your host\n# harness; this command runs the adversary panel + writes signed findings\n# to .tribunal/ledger.jsonl).\nANTHROPIC_API_KEY=sk-... tribunal review --plan P-42\n\n# Inspect what's in the ledger.\ntribunal ledger summary\ntribunal ledger leaderboard\n\n# v0.3: settle to Burnt XION. Deploy once, then sync per plan.\n./scripts/deploy-contract.sh                  # produces a chain.yaml snippet\ntribunal chain init --chain-id xion-testnet-2 --contract cosmwasm1... ...\ntribunal chain register claude-adversary\ntribunal chain sync --plan P-42\ntribunal chain query leaderboard\n```\n\n## Status\n\nThis repo is in active development. v0.1 ships the methodology, CLI, skills/agents, local ledger, and a Go fizzbuzz example. v0.2 adds multi-model adversarial dispatch and the real verification pyramid. v0.3 adds the CosmWasm contract and on-chain settlement. v0.5 adds the temporal lens, trajectory PBT, and cross-plan findings. v0.5.8 makes the reproducibility field on every finding mandatory (exploit path / trigger sequence / workload+numbers / manifesting cycle / PoC) so downstream maintainers can distinguish real threats from style violations in under 30 seconds.\n\nSee [`CHANGELOG.md`](./CHANGELOG.md) for what's released, and [`docs/methodology.md`](./docs/methodology.md) for the design.\n\n## Public site\n\n[**tribunal.mabus.ai**](https://tribunal.mabus.ai) — landing, the methodology rendered with sidebar nav, the [P-v032-audit case study](https://tribunal.mabus.ai/audits/p-v032-audit) (Tribunal reviewing its own release), and a [live on-chain leaderboard](https://tribunal.mabus.ai/leaderboard) that queries the deployed contract on `xion-testnet-2` client-side. Source under [`site/`](./site/).\n\n## Why Tribunal\n\nLLMs are fast, broad, and characteristically unreliable. Code review, tests, and audit are human-bottlenecked and scale linearly with reviewer attention. LLM output scales 10–100× faster. That mismatch is the trust gap.\n\nMost multi-agent systems being built today default to cooperative patterns: agents that help, vote, and converge. _That is the wrong primitive for correctness._ Cooperation amplifies shared mistakes. Adversaries hunt them.\n\nBut adversaries that are never held accountable hallucinate findings as freely as cooperators hallucinate code. Tribunal's bet: trust is a function of three things — surviving adversarial scrutiny, by agents with verifiable identity, whose history of findings is on the public record.\n\n## Documents\n\n- [Methodology](./docs/methodology.md) — the load-bearing design doc\n- [Incentive mechanism](./docs/incentive-mechanism.md) — reputation math\n- [On-chain protocol](./docs/on-chain-protocol.md) — CosmWasm contract surface (v0.3)\n- [Installation](./docs/installation.md) — per-host setup\n- [ADRs](./docs/adr/) — architecture decisions\n\n## License\n\n[GNU AGPLv3 or later](./LICENSE). Open-source, copyleft for network use — anyone running this as a service must publish their modifications under the same license.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fdpdanpittman%2Ftribunal","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fdpdanpittman%2Ftribunal","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fdpdanpittman%2Ftribunal/lists"}