skills.rest
CivicActions
CivicActionsOFFICIAL
Explore all skills by this author →
WebsiteGitHub

evidence-check

Validate AI-generated technical claims using a structured evidence taxonomy.

2|Updated May 24, 2026
Search Tags:#architecture#quality assurance#debugging#verification#evidence#honesty
One-click install
npx skills add https://github.com/CivicActions/ai-runbook-jh --skill evidence-check-civicactions
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: evidence-check
Source: https://github.com/CivicActions/ai-runbook-jh/tree/main/skills/evidence-check
Command: npx skills add https://github.com/CivicActions/ai-runbook-jh --skill evidence-check-civicactions

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill prevents AI from presenting guesses or hallucinations as established facts by forcing a rigorous verification process for all technical claims and recommendations.

Core Features & Use Cases

  • Evidence Taxonomy: Categorizes claims into tiers ranging from Verified (command output) to Unverified (training data patterns).
  • Active Auditing: Allows users to explicitly invoke an audit of previous AI responses to identify unsubstantiated claims.
  • Use Case: Use this skill when the AI recommends an architectural change or a specific library to ensure the suggestion is based on project-specific evidence rather than generic training data.

Quick Start

Invoke the evidence check skill to audit the previous response and verify the technical claims made about the current codebase.

Frequently Asked Questions about evidence-check

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

FAQPage Schema
How do I verify AI-generated technical claims about my codebase?▼

To verify AI-generated technical claims, you audit the response by cross-referencing project-specific configurations and documentation to distinguish between verified facts and probabilistic inferences. This structured evidence taxonomy enforces AI honesty during software development.

What is an evidence taxonomy for AI architectural recommendations?▼

An evidence taxonomy for AI architectural recommendations categorizes claims into tiers ranging from Verified, using command outputs, to Unverified, based on training data patterns. This mechanism prevents hallucinations by forcing rigorous verification.

Can I audit previous AI responses for unsubstantiated bug diagnoses?▼

Yes, you can audit previous AI responses for unsubstantiated bug diagnoses by explicitly invoking an active auditing process. This identifies claims not supported by project-specific evidence and enforces factual accuracy.

When do I need to enforce honesty for AI implementation suggestions?▼

You need to enforce honesty for AI implementation suggestions when the AI recommends architectural changes or specific libraries. This ensures suggestions are based on project-specific evidence rather than generic training data patterns.

Related Skills

Explore More Software Engineering Skills→
albibennialbibenni
2

project-onboarding

Analyze codebase structure and documentation to generate architectural summaries and quizzes.

Community
Intermediate
highflame-aihighflame-ai
4

from-issue

Transform GitHub issues into merge-ready pull requests with automated pipelines.

Official
Advanced
flet-devflet-dev
16.4k

run-integration-tests

Execute integration tests for the Flet project using README instructions.

Official
Basic
yongwoonyongwoon
8

ywc-receive-review

Verify code review feedback against the codebase before implementing changes.

Community
Advanced