Agent Engineering Toolkit

Proof-carrying workflows for coding agents

18 followers

Proof-carrying workflows for coding agents

18 followers

AET binds test runs, changed files, artifacts, and agent claims to exact code. Its local stale-proof demo shows a real PASS becoming inapplicable after relevant source changes—without an LLM, API key, cloud account, product telemetry, or holistic trust score.

Free

Launch tags:Open Source•Developer Tools•Artificial Intelligence

Launch Team

Framer AI AgentsDesign and publish professional sites with AI

Promoted

Maker

📌

I built AET after repeatedly seeing coding-agent handoffs where “tests passed” was true at execution time but no longer applied after later edits. The v1.18.0 stale-proof demo makes that boundary reproducible: it runs a real unittest, records EXACT_MATCH, changes relevant source, then reports RELEVANT_FILES_CHANGED while preserving the historical PASS. AET is local and evidence-first. It does not replace CI, infer complete coverage, score developer trust, or authorize a merge or release. I’d especially value feedback from maintainers who review agent-generated changes: which freshness errors create the most friction in real repositories?

Report

3mo ago

Ran the demo and watched a pass mark flip to inapplicable just by editing a few lines. No API keys or cloud setup was needed. Pretty cool to see reliability tracking that actually ties back to the exact code change.

Report

3mo ago

Reviews