跳到正文
TakoAPI
返回技能列表

clarity-gate

由 frmoretto 提供

Search & Research
337下载量由 clawskills.sh 提供
0浏览量
0 个赞登录后即可为这个技能点赞
—暂无评分
你的评分

简介

Pre-ingestion verification for epistemic quality in RAG systems.

描述

该技能的功能

Pre-ingestion verification system for RAG knowledge bases. Checks documents for missing uncertainty markers, unhedged claims, and implicit assumptions before they enter AI knowledge stores. Produces Clarity-Gated Documents (CGD) with a structured HITL verification record and validation error codes.

Unlike detection tools that find hedges already present in text, it flags where uncertainty markers are absent but required, catching hallucination risks at their source before documents enter knowledge bases.

何时使用

  • Checking a product spec before loading it into a RAG pipeline
  • Flagging unverified revenue projections in a planning document
  • Reviewing methodology docs for assumptions stated as facts
  • Validating a Source of Truth file before sharing across AI sessions
  • Auditing technical reports with specific statistics or competitor claims before publishing

示例工作流

输入

User asks: Verify this API pricing doc before we ingest it into our knowledge base

智能体
  1. 1

    Run 9-point epistemic check across all claims in the document

  2. 2

    Flag critical issues where forward-looking statements lack uncertainty markers

  3. 3

    Identify externally verifiable claims such as pricing figures and statistics for HITL routing

  4. 4

    Produce CGD frontmatter with clarity-status, hitl-pending-count, and per-claim records

  5. 5

    Present Round A and Round B verification requests for human confirmation

输出

A .cgd.md file with clarity-status: PENDING, a structured hitl-claims list, and a verification report categorized by Critical, Warning, Temporal, and Verifiable issues

安装

clawhub install frmoretto/clarity-gate

运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。

链接