TAMPER
SIGNAL

Documentation

Learn Tamper Signal.

A signed receipt at every step of your data pipeline, and one light that tells you whether the numbers on the dashboard still descend from the source. This is the deep version: the mental model first, then a hands-on quickstart, then full reference.

Start here

A three-step path

New to receipts and chains? Read these in order. Twenty minutes end to end, and you will understand both what the tool proves and how to wire it up.

  1. Core concepts

    The receipt, the chain, the two fingerprints, the traffic light. Explained with seals, notaries, and chains of custody before any code.

  2. Quickstart: your first verified chain

    Install, scaffold keys, ingest a source file, wrap a transform, and watch the light go green, then red when you tamper.

  3. Put the signal in your dashboard

    Mount the inline status light, the data tab, or the badge so anyone looking at the page can see the chain is intact.

Reference

Look something up

Concepts

The mental model

Receipt, chain, evidence vs semantic hash, control totals, the light, anchors, and the one claim it will never make.

Reference

CLI

Every receipts and tamper-signal command and flag, the exit codes, and the JSON verdict schema.

Reference

Python & JS API

Wrap transforms with receipt_step / receiptStep, ingest, rebuild, and verify in code.

Reference

Mounting the signal

Badge, status light, data tab, console, framework attach helpers, and flagging a broken metric in place.

Going deeper

Power user

Key rotation, CI signing, anchoring to a public log, cross-stack chains, drift, and format-agnostic hashing.

Hands-on

Quickstart

From pip install to a green light in about five minutes, then a deliberate tamper to see red.
For coding agents

Pointing an AI assistant at this project? These pages are written to be read by agents too, but the canonical integration runbook is AGENTS.md in the repo, with a compact map in llms.txt. Tell your agent "add tamper signal" and it will find them.