Aftermath

Trust the evidence, not the completion message.

Independent verification for agent-written code. Aftermath runs the gates and leaves a durable receipt.

How it works

The agent finishes. Aftermath investigates.

Claims are cheap. Exit codes, diffs, and baselines are not. Aftermath turns verification into evidence you can inspect, compare, and reuse in repair loops.

Agent writes

Cursor (or any agent) changes the repository and reports completion.

Aftermath verifies

Diff analysis, discovered commands, builds, tests, lint, typecheck, smoke, and policy checks — executed locally.

Evidence remains

A receipt in Markdown, HTML, and JSON. Findings with codes. Repair context when gates fail.

The product

A receipt proving what was actually verified.

Not an AI score. Not a vibes check. A deterministic verdict with preserved logs under .aftermath/.

AFTERMATH RECEIPT #184 LOCAL · NO TELEMETRY

VERDICT
PARTIALLY VERIFIED

BUILD     PASS
TESTS     941 passed · 4 failed
LINT      2 new warnings
API       1 potential break

Evidence: .aftermath/runs/0184/
Next:     aftermath repair-context latest

Start

Make verification the last step of every substantial session.

Link the Cursor plugin, create a baseline, then run /aftermath-verify when the agent says it’s done.

Install Read the docs