evidence-gate

Execute a fresh command to verify claims and return a PASS/FAIL verdict with output.

Updated May 21, 2026
One-click install
npx skills add https://github.com/random1st/groundwork --skill evidence-gate-random1st
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: evidence-gate
Source: https://github.com/random1st/groundwork/tree/main/plugins/evidence-gate
Command: npx skills add https://github.com/random1st/groundwork --skill evidence-gate-random1st

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Refuse completion without fresh verification by requiring an explicit command to prove each claim and presenting the resulting output as evidence.

Core Features & Use Cases

  • Enforce evidence before assertions by requiring a fresh run of a verification command.
  • Produce a PASS/FAIL verdict with the exact command output as supporting evidence.
  • Complement calibration tagging workflows to increase trust and traceability in AI reasoning.

Quick Start

Ask the agent to verify any claim with a fresh command and report PASS or FAIL with the exact output as evidence.

Frequently Asked Questions about evidence-gate

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

FAQPage Schema
How do I enforce command verification before an AI agent completes a task?

Enforce command verification by requiring a fresh command execution to prove each assertion before completion. The system runs the verification command and presents the exact output as evidence to gate the task.

What is claim gating in AI safety and how does it work?

Claim gating in AI safety is the process of refusing task completion without fresh proof. It works by executing a verification command provided by the user and returning a PASS or FAIL verdict with the exact output.

How do I get a pass-fail verdict with exact command output for test runs?

Get a pass-fail verdict by asking the agent to run a fresh verification command for your test or build checkpoint. The system executes the command and reports PASS or FAIL alongside the exact command output as supporting evidence.

Does evidence gating work for build verification and data-generation checkpoints?

Evidence gating works across tasks requiring verifiable output, including build verification and data-generation checkpoints. It requires a fresh command execution to prove the assertion and returns a PASS/FAIL verdict with the exact output.

Can I use evidence gating alongside calibration tagging workflows?

You can use evidence gating to complement calibration tagging workflows. It increases trust and traceability in AI reasoning by requiring fresh proof and returning a PASS/FAIL verdict with exact command output as evidence.

What are the limitations of requiring fresh proof for every claim?

Limitations of requiring fresh proof include the necessity of providing an explicit verification command for each assertion. The system cannot return a verdict without executing a fresh command and presenting the exact output as evidence.