Free Agent Skill Kit

Stop junk mail with an agent that asks before it acts.

A complete, open workflow for identifying who sent a piece of unwanted mail, finding their real opt-out process, and drafting a minimal request — with a hard rule baked into every layer: nothing gets sent until a person explicitly approves the exact request.

.zip · Markdown + JSON Schema · no account required

What this is

A reference implementation of the reasoning layer — eight Agent Skills, the JSON contracts they pass between each other, and the architecture doc describing the backend that has to enforce it. Deploy it yourself, point it at your own verified sender list, and it drafts real requests for real mail.

What this isn't

Not a hosted service, not a pre-loaded mailer database, and not legal advice. There is no universal “stop all mail” switch in U.S. law — most suppression is a voluntary sender preference, and a few narrow statutes (prescreened credit offers, certain USPS forms, sweepstakes removals) have their own official channels this kit routes to instead of automating.

Six steps, one hard gate.

Steps 1–3 and 5–6 can run unattended. Step 4 never does.

1

Capture

Photograph the sender's name, logo, or return address on the mailpiece.

2

Identify

The agent proposes a likely sender from a maintained registry — you confirm it, every time.

3

Draft

A minimal, channel-specific opt-out request is prepared and frozen — only the fields that sender actually requires.

4

You approve

You see the exact request on an authenticated page before anything sends. No approval, no send — ever.

5

Send

The request goes out through the sender's own accepted channel: email, form, postal, or fax.

6

Monitor

The case tracks the sender's own timeline and tells you honestly whether mail actually stopped.

The rule that makes this safe to build on.

What's in the download.

34 files. All Markdown and JSON Schema — readable in any editor, no build step required to start reading.

8 Agent Skills

manage-mail-preferences, inspect-mailpiece, resolve-mailer, research-optout-protocol, prepare-optout-request, authorize-optout-request, dispatch-optout-request, monitor-optout-case

4 operator-only skill stubs

Registry curation, protocol health, batch reconciliation, incident review — kept out of any consumer session

10 JSON Schema contracts

Every model ↔ backend boundary, from mailpiece evidence to the approval receipt

Architecture doc

The MCP tool surface, the approval → dispatch protocol, and the case state machine

Legal route matrix

Voluntary suppression vs. FCRA prescreening, USPS statutes, HIPAA, GLBA, and state privacy routes — with what an agent may and may not do on each

Phased roadmap

From an analyze-only prototype to a production service, with a launch checklist

Read before you deploy

This is a product and engineering blueprint, not legal advice, and it ships with zero pre-verified senderson purpose. Before you accept anyone's real mail, images, or address, have counsel review your authorization language, privacy notice, and route coverage — several of the statutory routes referenced in the kit require the person themselves, not an app, to sign or verify identity through an official channel. Full detail is in DISCLAIMER.md in the download.

Get the kit.

Free, no email required. Start with the analyze-only mode — it reads and drafts, never sends, and needs no backend at all.

Download the Starter Kit ↓

Want it built and hosted for you instead? Book a call.