---
schema_version: "newruntime-agent-readable-v0.1"
type: "project"
slug: "verifiable-ai-coding-workflow"
title: "Verifiable AI Coding Workflow"
description: "A small agent-assisted coding workflow where the useful artifact is not generated code, but a checked path from idea to tests, diff, review, and handoff."
status: "published"
project_status: "candidate"
one_line_outcome: "Turn an implementation idea into a bounded coding task with checks and reviewable evidence."
wow_moment: "The agent does not just produce a patch; it leaves commands, changed files, acceptance checks, and a human-readable handoff."
buildability_class: "B2"
bottleneck: "Requires a repo with runnable checks and a human willing to enforce task boundaries."
time_to_first_result: "1-2 hours in a prepared repository"
estimated_cost: "Low to medium, depending on model and test loop length"
code_level: "Developer or technical operator"
data_touched: "Repository source, tests, local command output"
write_actions: "Local file edits only; commits and deploys remain explicit approvals"
last_verified: "2026-07-21"
related_shift: "goal-scoped-agent-loops"
related_posts: ["ai-coding-workflow-verifiable-work","coding-agent-cost-environment-config"]
topics: ["coding-agents","developer-workflow"]
source_urls: ["https://github.com/mattpocock/skills"]
featured: true
routes: {"html":"https://newruntime.com/projects/verifiable-ai-coding-workflow/","markdown":"https://newruntime.com/projects/verifiable-ai-coding-workflow.md","json":"https://newruntime.com/projects/verifiable-ai-coding-workflow.json"}
source_format: "markdown"
---

# Verifiable AI Coding Workflow

## Minimal build

1. Write a task brief with files in scope, non-goals, and acceptance checks.
2. Give the agent repo-specific operating instructions.
3. Let it inspect before editing.
4. Require a focused diff and a real check command.
5. Keep commit, push, deploy, and external writes as separate approvals.

## Evidence

The current site update itself is a candidate use case: post content, docs,
production drift, and build checks had to be kept separate instead of collapsed
into one "make website better" prompt.