skills.rest
input-output-hk
input-output-hkOFFICIAL
Explore all skills by this author →
WebsiteGitHub

evidence-rules

Enforces citation and explicit-uncertainty rules for factual claims in agent outputs.

1.2k|296|Updated Oct 5, 2016
Search Tags:#fact-checking#verification#evidence#citation#sources#quotes#agent-rules
One-click install
npx skills add https://github.com/input-output-hk/daedalus --skill evidence-rules
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: evidence-rules
Source: https://github.com/input-output-hk/daedalus/tree/main/.agent/skills/evidence-rules
Command: npx skills add https://github.com/input-output-hk/daedalus --skill evidence-rules

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Agents often state unverified claims with false confidence or omit sources, making outputs hard to trust and audit. This Skill defines shared evidence rules so every factual claim is cited, unknowns are stated explicitly, and exact wording is quoted directly.

Core Features & Use Cases

  • Mandatory Citations: Requires factual claims to be backed by references to files, artifacts, or URLs.
  • Explicit Unknowns: Instructs agents to state when something is unknown or unverified instead of implying confidence.
  • Direct Quoting: Requires direct quotes when exact wording matters, such as legal, specification, or API text.
  • Use Case: When an agent reviews a codebase or writes a technical report, it cites the exact file or URL behind each claim and flags anything it could not verify.

Quick Start

Apply the evidence rules so that every factual claim in your answer includes a citation and any unverified points are stated explicitly.

Frequently Asked Questions about evidence-rules

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

FAQPage Schema
How do I make an AI agent cite its sources?▼

Apply evidence rules that require every factual claim to include a citation to a file, artifact, or URL. The agent must attach a verifiable reference instead of stating claims without support.

How to prevent AI agents from hallucinating facts?▼

Require the agent to state explicitly when something is unknown or unverified rather than implying confidence. Combining this with mandatory citations forces claims to be grounded in checkable sources.

When should direct quotes be used in agent outputs?▼

Use direct quotes when exact wording matters, such as specifications, legal text, or API definitions. Paraphrasing is acceptable for general claims, but precise language must be quoted verbatim from the source.

What counts as a valid citation for a factual claim?▼

A valid citation points to a concrete file, artifact, or URL that supports the claim. Vague references without a traceable source do not satisfy the evidence rules.

Can these evidence rules be combined with other agent instructions?▼

Yes, the rules are domain-agnostic and act as a baseline policy layered onto other skills. They govern how claims are presented, not what task the agent performs.

Related Skills

Explore More Content & Communication Skills→
andreaswasitaandreaswasita
33

cite-your-sources

Link every factual claim to an accessible source artifact.

Community
Intermediate
180VRisLife180VRisLife

post-referee

Resolves code-review disputes by inspecting cited source files and producing findings reports.

Community
Advanced
jikig-aijikig-ai
12

fact-checker

Extract verifiable claims from drafts and check cited URLs for supporting evidence.

Official
Intermediate
jhl-labsjhl-labs
50

RAG & Vector Search

Retrieve relevant documents and augment LLM prompts with vector search.

Official
Advanced