{"id":50804654,"url":"https://github.com/zhjai/groundcheck","last_synced_at":"2026-06-12T23:36:30.842Z","repository":{"id":361484058,"uuid":"1254622938","full_name":"zhjai/groundcheck","owner":"zhjai","description":"Single-agent, evidence-grounded claim verification to catch LLM hallucinations — a pluggable fact-gate for agent-arena and any multi-agent system (CrewAI, AutoGen, LangGraph).","archived":false,"fork":false,"pushed_at":"2026-06-01T16:33:14.000Z","size":26,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-06-12T23:35:59.765Z","etag":null,"topics":["agent-skill","claim-verification","claude-code-skill","fact-checking","groundedness","hallucination-detection","llm","rag-evaluation"],"latest_commit_sha":null,"homepage":null,"language":null,"has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/zhjai.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":"2026-05-30T20:00:16.000Z","updated_at":"2026-06-01T16:36:39.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/zhjai/groundcheck","commit_stats":null,"previous_names":["zhjai/groundcheck"],"tags_count":2,"template":false,"template_full_name":null,"purl":"pkg:github/zhjai/groundcheck","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zhjai%2Fgroundcheck","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zhjai%2Fgroundcheck/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zhjai%2Fgroundcheck/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zhjai%2Fgroundcheck/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/zhjai","download_url":"https://codeload.github.com/zhjai/groundcheck/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zhjai%2Fgroundcheck/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34266915,"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":["agent-skill","claim-verification","claude-code-skill","fact-checking","groundedness","hallucination-detection","llm","rag-evaluation"],"created_at":"2026-06-12T23:36:30.126Z","updated_at":"2026-06-12T23:36:30.836Z","avatar_url":"https://github.com/zhjai.png","language":null,"funding_links":[],"categories":[],"sub_categories":[],"readme":"# GroundCheck\n\n\u003cp align=\"center\"\u003e\n  \u003cimg src=\"assets/banner.svg\" alt=\"GroundCheck — single-agent, evidence-grounded claim verification to catch hallucinations\" width=\"100%\"\u003e\n\u003c/p\u003e\n\n\u003cp align=\"center\"\u003e\n  \u003cstrong\u003eEnglish\u003c/strong\u003e · \u003ca href=\"README.zh.md\"\u003e中文\u003c/a\u003e\n\u003c/p\u003e\n\n\u003cp align=\"center\"\u003e\n  \u003cimg src=\"https://img.shields.io/badge/skill-groundcheck-blue\" alt=\"skill\"\u003e\n  \u003cimg src=\"https://img.shields.io/badge/version-0.1.2-informational\" alt=\"version\"\u003e\n  \u003ca href=\"https://github.com/zhjai/agent-arena\"\u003e\u003cimg src=\"https://img.shields.io/badge/companion-agent--arena-6b46c1\" alt=\"agent-arena\"\u003e\u003c/a\u003e\n  \u003ca href=\"LICENSE\"\u003e\u003cimg src=\"https://img.shields.io/badge/License-MIT-yellow.svg\" alt=\"MIT License\"\u003e\u003c/a\u003e\n\u003c/p\u003e\n\n\u003e **Single-agent, evidence-grounded claim verification** — catch fabricated facts by checking each claim against real evidence, not by asking more models to agree.\n\nGroundCheck takes content that already exists — a generated answer, a report, RAG output, docs, or code — extracts the verifiable claims, grounds each one in real evidence, and returns a per-claim verdict with citations plus an action: **keep / revise / retract / send back**.\n\n## Why single-agent (and why that matters)\n\nMulti-agent debate is good at fixing **overconfidence** — but it can *reinforce* a shared **hallucination**, because several models trained on similar data confirm the same wrong fact. GroundCheck deliberately does **not** use a panel: it grounds every claim in deterministic external evidence (tests, source, docs, web, retrieved context).\n\n- **GroundCheck** → treats **hallucination** (single-agent + evidence grounding)\n- [**agent-arena**](https://github.com/zhjai/agent-arena) → treats **overconfidence** (multi-agent debate)\n\nThey form one verification stack at two depths and interoperate via a shared [Claim Ledger](interop/claim-ledger.md).\n\n## What it produces\n\nFor each claim: an atomic, classified, dated verdict —\n`supported · partially_supported · refuted · unverifiable · outdated · needs_qualification` —\nwith cited evidence and a recommended action. Compound claims are decomposed so a true\nsub-claim can't launder a false bundled conclusion.\n\nThe ledger is **contestable, not authoritative**: every verdict is traceable, time-bounded,\nand can be reopened with stronger counter-evidence. The verifier can be wrong too.\n\n## Use as a fact-gate in any multi-agent system\n\nGroundCheck is a **generic, pluggable verification gate** — agent-arena, CrewAI, AutoGen,\nLangGraph, or any orchestrator can plug it via the [fact-gate contract](interop/fact-gate.md):\n\n```\neach agent answers independently\n   → PRE-DEBATE gate: groundcheck per answer\n       → refuted? send back to that agent (with evidence) before debate\n   → debate\n   → POST-DEBATE gate: re-check new/changed claims\n```\n\nCatching factual errors **before** debate is what stops a panel from reinforcing a shared\nhallucination. Send-backs carry **evidence, not conclusions**, and the original answer stays\nimmutable — so the source agent reasons independently instead of appeasing the checker.\n\n## When to use / not use\n\n**Use:** verify claims · check for hallucinations · \"are these citations / numbers / APIs real\" ·\nfact-check before publishing · RAG groundedness · as a fact-gate in a multi-agent flow.\n\n**Not for:** pure opinions / creative content · code logic already covered by tests ·\n\"which option is better\" decisions (that's agent-arena).\n\n## Install\n\n```bash\nnpx skills add zhjai/groundcheck -g -a claude-code\n```\n\nWorks with Claude Code, Codex, Cursor, OpenCode, and other Agent Skills hosts. Or copy\n`skills/groundcheck/` into your agent's skills directory.\n\n## Status\n\n`v0.1.0` preview. Not affiliated with any vendor. MIT licensed.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fzhjai%2Fgroundcheck","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fzhjai%2Fgroundcheck","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fzhjai%2Fgroundcheck/lists"}