# Agent Intake Packet: brief a new agent without handing over secrets

> A free kit for adding an agent to the ones you already run. Every agent that knows something writes one dated page in the same seven sections. A checker stops keys, passwords, stale pages, pasted chat logs and topics you never share. You sign off at your own terminal, and only then does the new agent get its standing context, with a limits card on top.

- URL: https://paperchasewebb.ai/bots/marketplace/items/agent-intake-packet-brief-a-new-agent-without-handing-over-secrets/
- Type: Prompt pack
- Category: Solo business
- By: PaperChaseWebb
- Works with: Claude Code, Codex, Grok Bot, Hermes (named in the listing); One Python script, templates, a fictional example and agent rules for each harness. Tested on Linux
- Price: Free USD
- Difficulty: Intermediate
- Setup time: 15 min
- Time to result: First checked intake in about an hour
- Format: Markdown (.md) file, agent-intake-packet-brief-a-new-agent-without-handing-over-secrets.md
- Version: 1.0, updated 2026-11-03
- License: Personal and commercial use.

## What you end up with

A new agent that starts from short, dated, owner-approved briefings and a limits card, with no secrets and no old approvals carried over.

## What you get

- The intake method: plan in three tiers (pull first, desk card, never pull), ask, check, approve, build
- intake.py: 10 packet checks, 5 limits card checks and 5 plan checks, with secrets shown masked, never printed
- A seven-section packet template and a limits card template with the required lines
- Owner sign-off that is refused inside an agent session, and a build that refuses any file changed after sign-off
- AGENTS.md, CLAUDE.md, a Grok Bot standing order and a Hermes skill, a fictional four-source example, and 31 tests

## Who it is for

Anyone adding an agent to a fleet who wants it briefed from the others without secrets, stale facts or old approvals.

## Tools needed

- Python 3.8 or newer: intake.py: ask, check, owner approval and build; standard library only
- Grok Bot, Claude Code, Codex or Hermes (optional): collects packets and runs the checks under AGENTS.md; never approves
- You need: Python 3.8 or newer
- You need: Two or more agents you already run

Cost to run: Free. Python is free.

## How it works

1. **Run the example.** Unpack the file and run python3 intake.py check example/plan.json --today 2026-11-03. Expected: intake OK: 0 errors.
2. **Plan the intake.** List the agents that know something in plan.json and give each a tier. Anything with logins, money or personal details is tier 3. Expected: check shows no T codes.
3. **Collect and check.** Send each source the request from intake.py ask, save the replies under packets/, and run check until it is clean. Expected: 0 errors; warnings read and handled.
4. **Approve and build.** Run intake.py approve yourself, then build. Give the new agent context/standing-context.md. Expected: Standing context with the limits card first and no tier 3 content.

## Sample prompt: Collect packets (Grok Bot, Claude Code, Codex, Hermes)

```
Follow AGENTS.md. For each tier 1 and tier 2 source in plan.json, run intake.py ask and send that request to the source only. Save the replies under packets/, run intake.py check, and tell me in two lines what is ready. Do not run approve.
```

One prompt.

## Free preview of agent-intake-packet-brief-a-new-agent-without-handing-over-secrets.md

````markdown
# Agent Intake Packet

**Brief a new agent from the agents you already run: short, dated packets, a limits card, and your sign-off before it reads a word. No secrets cross over.**

Version 1.0, 2026-11-03. By PaperChaseWebb Inc. Free. Works with Grok Bot, Claude Code, Codex and Hermes, or by hand. Python 3.8 or newer, no installs.

## The problem

A new agent starts cold. The quick fix is to paste it everything: old chats, notes, a folder of files. That is how a key ends up in a prompt, a private detail ends up in a draft, and a decision from last month gets treated as a decision for today. Or the new agent reads an old "yes" and acts on it.

This kit replaces the paste with a short intake. Every agent that knows something writes one page in the same shape. A checker stops secrets, stale pages and raw dumps. You sign off. Only then does the new agent get its standing context, with the limits card on top.

## Who it is for

Owners and operators running two or more agents (a chief of staff, desks for booking, content, support or books) who are adding another one, or moving an agent to a new role. Also useful when you switch harnesses and want the new one to start from the same facts.

Not a memory system and not a permissions system. It produces one reviewed briefing. Your tools still decide what each agent can reach.

## How it works

1. **Plan (10 minutes).** List every source agent in plan.json and give each a tier.

| Tier | Means | What happens |
| --- | --- | --- |
| 1 | 
````

## Test record

- Date: Nov 3, 2026
- What was run: tests/test_intake.py (31 tests) unpacked from the pack file and run fresh on Linux, Python 3.8, 3.9, 3.11, 3.12 and 3.13; intake.py check on the fictional example.
- Result: 31 passed, 0 failed on all five versions; each check code caught by its own test; approval refused in agent sessions; build refused after an edit.
- Not run: macOS or Windows. A live Grok Bot, Claude Code, Codex or Hermes session.

Copied from this item's dated test notes. What was not run stays in.

## Worth knowing

- The secret patterns catch common key and password shapes, not every possible secret. Read the packets before you approve.
- It checks that the limits card says the right things. Your tools and your yes enforce them.
- On 2026-11-03: 31 tests passed on Python 3.8, 3.9, 3.11, 3.12 and 3.13 on Linux, run from the files unpacked out of the pack. Agent rules guide; they were not loaded inside each harness for this release.

## Questions

**Is this a memory system?** No. It produces one reviewed briefing for one agent. Run it again when the agent changes role or the packets go stale.

**Can my agent approve the packets?** No. approve refuses inside a Claude Code, Codex, Hermes or Grok Bot session and without a real terminal.

**What stops a key from getting through?** The checker looks for common key, token and password shapes and long random strings, and refuses to approve or build while any are present. It shows them masked.

**What should I pair it with?** Desk Authority Map ($29) sets who may decide what across people and agents, and Cross-Harness Relay ($39) covers handing work between harnesses once the new agent is running.

## How to get it

Free sample. Download it directly at https://paperchasewebb.ai/bots/marketplace/items/agent-intake-packet-brief-a-new-agent-without-handing-over-secrets/, no signup.

Seller: PaperChaseWebb Inc., mgmt@paperchasewebb.com