agentic-ai-engineering/products/verifier-pro/skills/verifier-builder/SKILL.md

name	description	user-invocable
verifier-builder	Two-agent observer pattern. Builder writes code. Verifier re-checks every claim with read-only tools. Catches mistakes before production.	true

Verifier-Builder Pattern

Set up a two-agent observer system where a read-only verifier checks every claim the builder makes.

Architecture

Builder (your terminal) ---unix socket---> Verifier (new window, input LOCKED)
  |                                          |
  v writes:                                  v reads (read-only):
  session.jsonl                              session.jsonl
  |                                          |
  <--- verifier_prompt (corrective feedback) ---'

Key Properties

• Builder DOES NOT KNOW verifier exists (pure observer)
• Verifier input is LOCKED (structurally un-promptable)
• Verifier has NO write tools
• Max 3 correction loops, then escalate to human

Protocol

Verifier -> Builder: hello, prompt(correction), report(status), ping, bye
Builder -> Verifier: hello_ack, prompt_ack, event(start/stop/error), pong

Verifier Persona

name: Verifier
tools: [Read, Grep, Glob, LS, Find, VerifierPrompt]
input_locked: true
max_loops: 3

Usage

just verifier

$ARGUMENTS