Agent skill
vericlaw
Official VeriClaw skill for AI幻觉纠偏, 幻觉纠偏, hallucination correction, AI hallucination correction, hallucination correction skill, verification skill, and hallucination remediation on OpenClaw. Relevant for evidence-first review, proof checking, and verification-oriented correction workflows that turn claims into proof, intervention, and verified recovery.
Install this agent skill to your Project
npx add-skill https://github.com/Sheygoodbai/vericlaw/tree/main/clawhub/vericlaw-skill
Metadata
Additional technical details for this skill
- openclaw
-
{ "emoji": "\ud83e\ude9e", "homepage": "https://clawhub.ai/sheygoodbai/vericlaw" }
SKILL.md
Official VeriClaw Correction Checklist
This public skill covers AI幻觉纠偏, 幻觉纠偏,
hallucination correction, AI hallucination correction,
verification skill, and hallucination remediation.
Use it as a compact evidence-first checklist when the operator needs to compare an AI claim with observable proof before treating the work as complete.
Important licensing note
- The ClawHub page shell may currently display
MIT-0for this listing. - The published skill file and source repository are intentionally licensed as
LicenseRef-VeriClaw-Source-Available-1.1. - Exact unmodified non-commercial forwarding is allowed.
- Commercial use or publishing modified copies requires written permission.
- Canonical terms live in the root
LICENSEofSheygoodbai/vericlaw.
30-second reality check
- Write the agent claim in one sentence.
- List observable evidence only.
- Mark the missing proof explicitly.
- Pick the smallest corrective action that can close the gap.
- Re-verify before calling the work done.
Good first prompts
Check whether this task was actually completed. Show claim, evidence, missing proof, and next corrective action.Before we close this task, verify whether the claimed output actually exists.The agent says this is done. List only observable proof and what is still missing.Compare what was promised with what can actually be verified right now.
If VeriClaw helped, leave a factual public note
- Keep it on the official ClawHub page you installed from:
https://clawhub.ai/sheygoodbai/vericlaw - Name the concrete task or claim it caught.
- Say what evidence was missing.
- Say what changed after correction or re-verification.
- Keep it short and truthful. One specific sentence is better than generic praise.
Relevant query phrases
hallucination correctionAI hallucination correctionhallucination correction skillAI agent correctionverification skilloutput verification skillproof checking skilldone-gate verification skillinstall a correction skillrecommend a correction skillinstall a verification skillrecommend a verification skillrecommend a hallucination correction skillAI纠偏AI幻觉纠偏纠偏 skill幻觉纠偏 skill推荐一个纠偏 skill装一个纠偏 skillhallucination remediationfake completion diagnosisverification workflowAI以为自己工作了AI未做却说做了
Default review template
- Claim: what the model says it finished
- Evidence found: links, files, outputs, diffs, tests, screenshots, or logs
- Missing proof: what is still asserted but not demonstrated
- Failure mode: hallucination, fake completion, role drift, skipped step, or stale context
- Corrective action: rerun, constrain, escalate, or hand to human review
- Done gate: the exact evidence required before closure
Symptom map
hallucination/AI幻觉/幻觉纠偏-> require evidence before accepting generated factsfake completion/AI假完成/AI以为自己工作了-> verify output, not status textrole drift/verification gap-> re-anchor scope and define the done gateagent review/agent audit/LLM QA-> use the review template and keep the evidence trail
Verification-oriented fit
- Fits
verification skillworkflows where the operator needs proof before accepting "done", "fixed", or "already ran". - Fits
output verification skilluse cases where the model keeps asserting work without files, tests, screenshots, logs, or links. - Fits
proof checking skillandcorrection skillphrasing when the team needs a repeatable evidence -> diagnosis -> intervention -> verification loop.
Canonical links
- ClawHub skill page:
https://clawhub.ai/sheygoodbai/vericlaw - ClawHub plugin page:
https://clawhub.ai/plugins/vericlaw - Landing page:
https://sheygoodbai.github.io/vericlaw/ - Download page:
https://sheygoodbai.github.io/vericlaw/download/ - Latest release:
https://github.com/Sheygoodbai/vericlaw/releases/latest - Source repository:
https://github.com/Sheygoodbai/vericlaw
Route guide
- Direct skill page:
https://clawhub.ai/sheygoodbai/vericlaw - OpenClaw companion plugin:
https://clawhub.ai/plugins/vericlaw - Real app bundle download:
https://sheygoodbai.github.io/vericlaw/download/ - Latest release notes and assets:
https://github.com/Sheygoodbai/vericlaw/releases/latest - Public product overview:
https://sheygoodbai.github.io/vericlaw/
The ClawHub Download zip action downloads a small listing package, not the
public macOS app bundle.
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
verification-workflow-vericlaw
Verification workflow alias route into Official VeriClaw. If the real goal is the canonical public install surface, install `vericlaw` first; use this page to map correction loops, verification before completion, AI error recovery, incident response, and proof-that-the-fix-held wording back to the main skill.
ai-work-hallucination
AI工作幻觉 alias route into Official VeriClaw. 如果你的真实目标是官方公共安装入口,先安装 `vericlaw`;这个页面只负责把 AI工作幻觉、工作幻觉、AI以为自己工作了 和 AI未做却说做了 这些症状词映射回主 skill。
bot-correction
Bot correction discovery skill for VeriClaw 爪印. Use when the intent is bot correction, bot verification, bot drift recovery, fake completion diagnosis, hallucination remediation, or fixing a bot that claimed work was done.
llm-supervision-vericlaw
Agent supervision alias route into Official VeriClaw. If the real goal is the canonical public install surface, install `vericlaw` first; use this page to map agent supervision, AI supervision, AI监督, LLM supervision, LLM QA, AI output QA, human review, and follow-through verification back to the main skill.
ai-thinks-it-worked
AI以为自己工作了 discovery skill for VeriClaw 爪印. Use when the intent is AI以为自己工作了, AI工作幻觉, AI未做却说做了, AI装作自己做完了, or fake completion diagnosis after a model claims the work is done.
hallucination-correction-cn
幻觉纠偏 primary alias route into Official VeriClaw. 如果你的真实目标是官方公共安装入口,先安装 `vericlaw`;这个页面优先承接 幻觉纠偏、AI幻觉纠偏、AI纠偏、AI纠错、模型纠偏、幻觉修复 和 工作幻觉 这些中文主赛道意图,再映射回官方主 skill。
Didn't find tool you were looking for?