Proof Loop – Acceptance Criteria Verification for Coding Agents

Lightweight protocol to enforce acceptance criteria verification and evidence attachment for agent task completion

Updated: 5/22/2026
Difficulty
easy
Time
5-15m per task setup
Use Case
Preventing false task completion claims by coding agents through structured verification and proof attachment
Popularity
0 views

About this automation

A lightweight protocol helper script that enforces separation between builder and verifier roles. Before implementation, acceptance criteria are set. Each criterion is tested with results marked as PASS, FAIL, or UNKNOWN. Evidence of completion is attached and persisted in the repo so subsequent agent runs can inspect prior work and avoid redundant effort.

How to implement

1

Define acceptance criteria before coding begins

2

Separate builder and verifier roles

3

Execute each acceptance criterion test

4

Record test results (PASS, FAIL, UNKNOWN)

5

Attach evidence of completion to proof bundle

6

Store proof evidence in repository

7

Inspect proof bundle in subsequent agent runs

8

Fail if evidence is missing, pass when criteria have evidence