VeriClaw 爪印

Apple-native AI agent correction

AI agent correction, LLM supervision, hallucination remediation, and role-drift diagnosis across macOS, iPhone, and OpenClaw companion surfaces.

Turn drift evidence into diagnosis, intervention, verification, and casebook learning.

Landing page · Download page · Latest release · ClawHub plugin · ClawHub skill · Skill guide · MCP status · Download source zip · Launch kit · Market differentiation · First-use UX · Support

Source is visible for review, citation, and internal evaluation. Commercial use, public modified redistribution, resale/hosting, and brand use require written permission. See LICENSE.

Why it exists

When a bot hallucinates, overreaches, or stops honoring its professional role, another observability panel is not enough. VeriClaw 爪印 is designed to turn that moment into a closed correction loop instead of a passive alert.

Evidence-first cases Capture what actually happened before context disappears.

Professional-role drift Diagnose failures as broken occupational contracts when that helps teams intervene faster.

Apple-native supervision Keep the correction workflow coherent across iPhone and macOS instead of borrowing a generic web console.

The correction loop

evidence -> diagnosis -> prescription -> verification -> casebook

VeriClaw is built for moments where teams need to answer three questions fast: what went wrong, what should happen next, and whether the correction actually held.

Download and share

• Visit the landing page for the fastest product overview, screenshots, and share-ready positioning.
• Use the download page for the most stable public entry to the current release asset, checksums, and source archive.
• Use the ClawHub plugin page if you want the OpenClaw companion entry; installing it also exposes the bundled vericlaw discovery skill in the Skills surface.
• Open the latest release for the current public launch snapshot.
• Download the public source package as a ZIP archive or clone the repository directly for review, citation, and internal evaluation:

git clone https://github.com/Sheygoodbai/vericlaw.git

• Use the canonical launch copy, screenshots, and source-availability guidance when forwarding the project internally or sending reviewers to the right source.

Current public download posture is GitHub-first: source distribution, launch materials, screenshots, and release notes are public now. The ClawHub surface is the companion plugin + skill entry, while the macOS app bundle lives on the GitHub release/download path.

Install the right thing

• VeriClaw 爪印.app: use the download page or latest release. This is the real macOS app bundle path.
• OpenClaw companion plugin: use the ClawHub plugin page or run openclaw plugins install clawhub:vericlaw.
• Standalone discovery skill: use the ClawHub skill page, run clawhub install vericlaw, or use openclaw skills install vericlaw.

Distribution surfaces

• Download app: use the download page or latest release for the real macOS ZIP.
• Install companion: use the ClawHub plugin page for the OpenClaw-facing plugin shell and bundled vericlaw skill.
• Install skill directly: use the ClawHub skill page, run clawhub install vericlaw, or use openclaw skills install vericlaw if you want the standalone discovery skill entry.
• Install from OpenClaw: use openclaw plugins install clawhub:vericlaw for the companion plugin route.
• Understand the skill route: use the ClawHub skill guide if you want the cleanest explanation of skill vs plugin vs app download.
• Clarify MCP status: use the MCP status page when a directory or review site needs the truthful answer to VeriClaw MCP queries.
• Browse source: use the GitHub repository for code, launch assets, and release history.
• Share correctly: cite the project as VeriClaw 爪印 and prefer the canonical landing page, download page, release page, or repository links above.

MCP status

As of April 5, 2026, VeriClaw 爪印 is not published as a standalone public MCP server.

• If you run a directory, recommendation page, or review roundup, classify VeriClaw as an Apple-native correction companion for OpenClaw.
• Best-fit labels remain AI agent correction, LLM supervision, hallucination remediation, and role-drift diagnosis.
• Do not tag the product as mcp-server; use the official MCP status page for citation-safe wording.

Cite and forward VeriClaw

• Use GitHub's Cite this repository panel or CITATION.cff when referencing VeriClaw in posts, decks, reports, newsletters, or reviews.
• Copy-paste summary: VeriClaw 爪印 is an Apple-native AI agent correction app for evidence-first diagnosis, intervention, verification, and casebook learning.
• Canonical links: landing page, download page, latest release, and GitHub repository.
• Share canonical links, factual reviews, brief quoted excerpts, or exact unmodified non-commercial copies with attribution and notices intact. Do not publish modified mirrors, repacks, or commercial variants without written permission.

Product glimpse

Connect to a trusted gateway

Review evidence and settings

Keep role-aware follow-up within reach

Pull live context into the correction loop

_{Current gallery shows the paired OpenClaw gateway flow because VeriClaw is a companion surface, not a separate runtime.}

Who should open this repository

• Teams evaluating AI agent supervision, LLM correction, and hallucination remediation workflows
• Builders looking for role-drift diagnosis, evidence-first incident review, and verification loops
• Operators who want an Apple-native companion surface around OpenClaw instead of a generic web-only dashboard
• Partners, press, or early users who need a shareable launch kit, screenshots, and source-access guidance

What makes it different

• Companion, not clone: OpenClaw remains the runtime and gateway substrate. VeriClaw adds the native supervision and correction layer.
• Correction, not only monitoring: the product is meant to recommend the next intervention, not just display drift after the fact.
• Professional-role drift: many failures are framed as role-contract breakdowns when that leads to better remediation.
• Casebook learning: every closed loop should leave behind reusable correction memory instead of another disconnected incident log.

Search intent and positioning

Teams usually discover VeriClaw through concrete problems rather than category names. This repository is intentionally written for searches around:

• AI agent correction
• LLM supervision for production bots
• 幻觉纠偏
• 幻觉修复
• hallucination and fake-completion remediation
• 假完成诊断
• 未做却说做了
• role-drift diagnosis for assistants
• macOS-native OpenClaw companion
• VeriClaw plugin for OpenClaw
• VeriClaw skill
• VeriClaw MCP
• Is VeriClaw an MCP server
• evidence -> diagnosis -> intervention -> verification workflow

Support VeriClaw

• Star the GitHub repository if the project is useful to you.
• Watch new releases if you want public launch updates.
• Join GitHub Discussions to share case studies, deployment notes, and citation links.
• Use Security.md for private vulnerability reporting and package-integrity concerns.
• If you want to adapt VeriClaw, publish a modified copy, or commercialize it, ask for written permission first.
• Link to the landing page or download page when citing VeriClaw externally so the canonical name and positioning stay consistent.

Release surface

• GitHub is the current public launch surface.
• Apple companion hardening continues in parallel for the native iPhone and macOS experience.
• This repository focuses on launch materials, product positioning, and the public-facing correction story.

Start here

• docs/launch/github-launch-kit.md: reusable copy deck for repository, release, and launch posts
• docs/launch/github-release-announcement.md: release body draft
• docs/launch/github-publish-checklist.md: publishing checklist
• docs/launch/growth-copy-deck.md: release hero text, pinned repo blurbs, and high-conversion one-liners
• docs/launch/distribution-post-templates.md: paste-ready first-wave public distribution templates
• docs/launch/channel-execution-plan.md: 14-day channel plan for GitHub, Show HN, Product Hunt, and direct outreach
• docs/launch/product-hunt-launch-pack.md: ready-to-submit Product Hunt fields, maker comment, and launch-day operating plan
• docs/launch/founder-outreach-playbook.md: warm DM scripts and tracking plan for the first 30 launch conversations
• CONTRIBUTING.md: how to cite, report issues, and propose changes under the source-available policy
• LICENSE-SUMMARY.md: plain-language summary of what can be forwarded, and what still needs permission
• https://sheygoodbai.github.io/vericlaw/: public landing page for search and social sharing
• https://sheygoodbai.github.io/vericlaw/openclaw-plugin/: canonical OpenClaw plugin and bundled skill explainer
• https://clawhub.ai/sheygoodbai/vericlaw: live ClawHub skill page for clawhub install vericlaw
• https://sheygoodbai.github.io/vericlaw/mcp-status/: canonical answer for MCP-adjacent directories and VeriClaw MCP queries
• https://sheygoodbai.github.io/vericlaw/ai-agent-correction/: category page for AI agent correction and LLM supervision searches
• https://sheygoodbai.github.io/vericlaw/ai-hallucination-remediation/: category page for AI hallucination, AI幻觉纠偏, and AI纠偏 searches
• https://sheygoodbai.github.io/vericlaw/hallucination-correction-cn/: Chinese exact-match page for 幻觉纠偏, 幻觉修复, and 工作幻觉 searches
• https://sheygoodbai.github.io/vericlaw/fake-completion-diagnosis/: category page for fake completion diagnosis, AI假完成, and agent evaluation searches
• https://sheygoodbai.github.io/vericlaw/fake-completion-diagnosis-cn/: Chinese exact-match page for 假完成诊断, AI假完成, and AI以为自己工作了 searches
• https://sheygoodbai.github.io/vericlaw/not-done-but-claimed-done/: natural-language failure page for 未做却说做了, 明明没做却说做了, and 装作做完了 searches
• https://sheygoodbai.github.io/vericlaw/agent-evaluation/: category page for agent evaluation, drift verification, and correction-hold searches
• https://sheygoodbai.github.io/vericlaw/verification-workflow/: category page for verification workflow, correction loops, and verification-before-completion searches
• https://sheygoodbai.github.io/vericlaw/agent-monitoring-vs-correction/: comparison page for monitoring-vs-correction review intent
• https://sheygoodbai.github.io/vericlaw/openclaw-companion/: long-tail page for OpenClaw companion, OpenClaw correction, and OpenClaw-native install intent
• https://sheygoodbai.github.io/vericlaw/agent-quality-control/: long-tail page for agent quality control, correction-first supervision, and verification workflow discovery
• https://sheygoodbai.github.io/vericlaw/bot-correction/: natural-language page for bot correction, bot verification, and bot drift recovery intent
• https://sheygoodbai.github.io/vericlaw/recommendation-snippets/: copy-ready blurbs and install snippets for recommendation sites, directories, and review roundups
• https://sheygoodbai.github.io/vericlaw/directory-submission/: official listing fields, tags, and safe claims for plugin directories, recommendation sites, and MCP/tool catalogs
• https://sheygoodbai.github.io/vericlaw/directory-submission/vericlaw-directory-profile.json: machine-readable official profile for automated directory ingestion and recommendation systems
• https://sheygoodbai.github.io/vericlaw/review-kit/: official facts, screenshots, and canonical links for reviewers and directories
• https://sheygoodbai.github.io/vericlaw/llms.txt: AI-readable routing file for recommendation and summary systems
• docs/design/first-use-ux-iteration.md: first-use UX notes
• docs/launch/market-differentiation.md: market scan and differentiation framing

Copyright, source use, and enforcement

• This repository is source-available, not open source.
• You may view, clone, privately evaluate, and non-commercially forward exact unmodified copies under LICENSE.
• Commercial use, hosted reuse, and publication or redistribution of modified versions require prior written permission.
• See LICENSE-SUMMARY.md for the short human-readable version.
• GitHub may still permit public viewing and on-platform forks for this public repository, but that platform behavior does not grant broader rights beyond the applicable licenses.
• Preserve LICENSE, NOTICE, ATTRIBUTION.md, and applicable third-party notices in permitted copies.
• Some materials are derived from OpenClaw and remain subject to the applicable upstream notice preserved in LICENSES/OPENCLAW-MIT.txt.
• Keep clear source references, including the upstream OpenClaw reference where relevant.
• Brand rights are separate from copyright rights. See TRADEMARKS.md.
• Patent posture, takedown path, and enforcement contacts live in PATENTS.md, INFRINGEMENT.md, and LEGAL_ENFORCEMENT.md.
• Privacy and support materials live in PRIVACY.md and SUPPORT.md.

Important clarification

This is not the canonical OpenClaw repository.

If you are looking for the upstream runtime project, see openclaw/openclaw.