skills.rest
Picrew
PicrewCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

rag

Answer questions using only prompt material with evidence-backed citations.

32|Updated Mar 2, 2026
Search Tags:#rag#compliance#policies#evidence-units#cited-answer#prompt-only
One-click install
npx skills add https://github.com/Picrew/OpenRAG-Skill --skill rag-picrew
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: rag
Source: https://github.com/Picrew/OpenRAG-Skill/tree/main
Command: npx skills add https://github.com/Picrew/OpenRAG-Skill --skill rag-picrew

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes assets (resource) and references (resource) components.

What problem does it solve?

It enables answering questions using only material provided in the prompt, producing a cited, evidence-backed result, and it refuses to guess when material is missing.

Core Features & Use Cases

  • Evidence Units: turns the provided text into stable, citeable evidence IDs.
  • Two-pass workflow: first locate evidence, then draft the answer.
  • Normative force lock: preserves SHALL, SHOULD, and MAY distinctions.
  • Output contract: yields a consistent answer structure for review and scoring.
  • Refusal discipline: uses INSUFFICIENT_EVIDENCE when material is not sufficient.
  • Use cases: policy interpretation, contract comparison, audits, and compliance checks grounded in the provided material.

Quick Start

Paste your source material into the chat and run the $rag prompt to produce a cited, evidence-backed answer.

Frequently Asked Questions about rag

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I get cited answers from policy documents using only provided context?▼

To get cited answers from policy documents, paste your source material into the chat and run the prompt. The system packages text into Evidence Units with stable IDs, then locates evidence and drafts a cited answer in two passes.

How do I handle contract comparison when the provided material is missing clauses?▼

When material is insufficient for contract comparison, the system applies refusal discipline and outputs INSUFFICIENT_EVIDENCE rather than guessing. It only answers when the provided context contains enough evidence units to ground the response.

Does the evidence-first Q&A process preserve SHALL, SHOULD, and MAY normative distinctions?▼

Yes, evidence-first Q&A applies a normative force lock that preserves SHALL, SHOULD, and MAY distinctions from the provided material. This ensures compliance interpretations maintain exact clause boundaries and regulatory force from the original policy text.

Can I use this for extraction and audit tasks on already loaded long-form content?▼

Yes, you can use this for extraction, comparison, and audit tasks on long-form content already loaded in context. The system requires packaging context into Evidence Units while preserving structure and clause boundaries to produce grounded, cited answers.

What are the limitations of prompt-only evidence-backed Q&A for policy interpretation?▼

The limitation of prompt-only Q&A is that it answers strictly from material provided in the prompt and refuses to answer when evidence is insufficient. It cannot retrieve external information or guess beyond the loaded context boundaries.

Related Skills

Explore More Legal & Compliance Skills→
0xsline0xsline
907

long-video-to-shorts

Identify and extract high-engagement short-form video segments from long-form source material.

Community
Advanced
MoMarcode1MoMarcode1
1

recht-review

Automate Austrian contract reviews with multi-agent clause analysis and RIS verification.

Community
Advanced
rovannirovanni
1

docx

Create and edit Word .docx documents with formatting and embedded content.

Community
Advanced
haxlyshaxlys

rag-implementation

Build Retrieval-Augmented Generation systems with vector stores and semantic search.

Community
Advanced