home / skills / sickn33 / antigravity-awesome-skills / clarity-gate

clarity-gate skill

This skill helps ensure epistemic quality in RAG pipelines through pre-ingestion checks and a 9-point verification with two-round HITL.

This is most likely a fork of the clarity-gate skill from xfstudio

npx playbooks add skill sickn33/antigravity-awesome-skills --skill clarity-gate

Review the files below or copy the command above to add this skill to your agents.

Files (1)

SKILL.md

838 B

---
name: clarity-gate
description: "Pre-ingestion verification for epistemic quality in RAG systems with 9-point verification and Two-Round HITL workflow"
source: "https://github.com/frmoretto/clarity-gate"
risk: safe
---

# Clarity Gate

## Overview

Pre-ingestion verification for epistemic quality in RAG systems with 9-point verification and Two-Round HITL workflow

## When to Use This Skill

Use this skill when you need to work with pre-ingestion verification for epistemic quality in rag systems with 9-point verification and two-round hitl workflow.

## Instructions

This skill provides guidance and patterns for pre-ingestion verification for epistemic quality in rag systems with 9-point verification and two-round hitl workflow.

For more information, see the [source repository](https://github.com/frmoretto/clarity-gate).

Overview

This skill provides a pre-ingestion verification layer to improve epistemic quality in retrieval-augmented generation (RAG) pipelines. It applies a 9-point verification rubric and a two-round human-in-the-loop (HITL) workflow to filter, score, and annotate content before it enters the knowledge store. The result is higher factual reliability and clearer provenance for downstream agents.

How this skill works

The skill inspects incoming documents and metadata across nine verification criteria covering provenance, source authority, temporal relevance, factual consistency, attribution, redundancy, bias indicators, format fidelity, and extraction confidence. It then runs a two-round HITL process: an initial reviewer applies the 9-point checklist and flags issues, and a second reviewer resolves disagreements or escalates uncertain items. Outputs include a verification score, standardized annotations, and remediation suggestions.

When to use it

Before ingesting large external corpora into a RAG vector store.
When regulatory or domain compliance requires documented verification steps.
To reduce hallucination risk in downstream generative responses.
When building high-stakes agents that require traceable evidence and source quality.
During dataset curation or periodic re-verification of knowledge bases.

Best practices

Automate the mechanical checks (format, timestamps, source resolution) and reserve human reviewers for judgment calls like bias and factual conflicts.
Calibrate the 9-point rubric with sample documents from your domain to ensure inter-rater reliability.
Log reviewer decisions, comment threads, and final scores for auditability and continuous improvement.
Use the two-round HITL to balance throughput and accuracy: quick triage first, deeper adjudication second.
Integrate remediation steps (reject, annotate, enrich, or link to primary sources) as structured outputs to feed ingestion workflows.

Example use cases

Curation of news articles and reports before adding them to a legal or medical RAG system.
Onboarding third-party knowledge feeds for an enterprise conversational assistant with audit requirements.
Periodic quality sweeps of an academic or research corpus to flag outdated or contradicted claims.
Pre-processing scraped web content to reduce low-quality or unverified facts from training and retrieval stores.

FAQ

What is the 9-point rubric based on?

The rubric covers provenance, authority, timeliness, factual consistency, attribution, redundancy, bias indicators, format fidelity, and extraction confidence to provide multifaceted epistemic assessment.

How much human effort does the two-round HITL require?

Effort depends on volume and automation: many mechanical checks can be automated, leaving human reviewers to focus on ambiguous or high-risk items; the two-round model prioritizes throughput with quality control.