---
title: "Agents Field Guide"
description: "Prompts, methods, and worked examples for building with agents."
---

> Documentation Index
> Fetch the complete documentation index at: https://afg.callum.website/llms.txt
> Use this file to discover all available pages before exploring further.

# Agents Field Guide

[Understand an existing product](https://afg.callum.website/understand-an-existing-product/) · [Map a flow](https://afg.callum.website/breadboard-to-proof/) · [Prove the result](https://afg.callum.website/prove-the-user-outcome/) · [Recover drifting work](https://afg.callum.website/recover-drifting-work/)

<!-- Generated catalogue -->

## [Skills](https://afg.callum.website/skills/)

- [Skills I publish](https://afg.callum.website/skills-i-publish/). My collection on skills.sh.
- [External skills](https://afg.callum.website/external-skills/). Agent coding methods, UI skills, and tools from other authors.

## [Prompts](https://afg.callum.website/prompts/)

- [Ask for proof of work instead of a summary](https://afg.callum.website/ask-for-proof-of-work/). Ask a worker for evidence you can inspect.
- [Find implicit state machines with Pstack](https://afg.callum.website/find-implicit-state-machines/). Model each subsystem and chart its key flows.
- [Review untested branches](https://afg.callum.website/prompts/review-untested-branches/). Decide which checks need tests and which states the types can rule out.

- [Could a human working for free do it?](https://afg.callum.website/prompts/outsource-to-a-human/) Test what it would take to hand off the work.
- [Restate the goal](https://afg.callum.website/prompts/restate-the-goal/). Check that the agent understood your goal before it starts.
- [Find a better method](https://afg.callum.website/prompts/find-a-better-method/). Reconsider the method you assumed the work required.
- [Design a system to do the work](https://afg.callum.website/prompts/design-a-system-to-do-the-work/). Reframe a task as a system to build.
- [What is worth building now?](https://afg.callum.website/prompts/what-is-worth-building-now/) Look for work that became practical as the cost changed.
- [Deeper without bigger](https://afg.callum.website/prompts/deeper-without-bigger/). Ask what better practice would add to the same task.
- [The best next question](https://afg.callum.website/prompts/best-next-question/). Get the question that would move the work forward.
- [The same with half the code](https://afg.callum.website/prompts/half-the-code/). Ask for a smaller implementation of the same thing.
- [Stay at the frontier](https://afg.callum.website/prompts/stay-at-the-frontier/). Resolve the decisions you can see now.

## [Playbooks](https://afg.callum.website/methods/)

- [Describe how an existing product behaves](https://afg.callum.website/understand-an-existing-product/). Trace what people experience when a flow is interrupted.
- [Take a breadboard from sketch to proof](https://afg.callum.website/breadboard-to-proof/). Prove an upload survives reopening.
- [Prove the thing the user can actually do](https://afg.callum.website/prove-the-user-outcome/). Find where to act and what would prove it worked.
- [Get an agent unstuck when it keeps missing the point](https://afg.callum.website/recover-drifting-work/). Find the first miss and choose a repair.
- [Work in independently valuable steps](https://afg.callum.website/independently-valuable-steps/). Use each delivery to pick the next step.
- [Turn repeatable agent work into a script](https://afg.callum.website/turn-repeatable-agent-work-into-a-script/). Decide when the steps are ready to automate.
- [Give the agent a way to verify its work](https://afg.callum.website/give-the-agent-a-way-to-verify/). Fast checks for flows and invariants.
- [Give a long task a worklog](https://afg.callum.website/give-a-long-task-a-worklog/). Keep decisions and open questions from the chat.
- [Coordinate work in Codex](https://afg.callum.website/coordinate-work-in-codex/). Split work across agents and chats, and keep a shared worklog.
- [Intent specialists](https://afg.callum.website/give-a-job-to-a-specialist/). Six reusable agent roles and when to use them.
- [Automate agreed rules and surface open decisions](https://afg.callum.website/automate-agreed-rules/). Bring open contract questions to review.

## [Field reports](https://afg.callum.website/field-reports/)

- [T3 Code with Opus and DeepSeek](https://afg.callum.website/t3-code-experiment/). *In progress.* Opus directs DeepSeek workers.
- [Why I turned off automatic agent memory](https://afg.callum.website/why-i-turned-off-automatic-agent-memory/). *Stopped.* Four tasks with and without access to past lessons.
- [AgentsView learning](https://afg.callum.website/agentsview-learning/). *Prob nah.* Search worked. Raw logs caught failures.

## [References](https://afg.callum.website/references/)

- [Four Dark Corners of Requirements Engineering](https://afg.callum.website/four-dark-corners-of-requirements-engineering/). From specification to outcome.
- [Towards Self-Driving Codebases](https://afg.callum.website/towards-self-driving-codebases/). Give agents a way to run and test their changes.
- [The Topology of an LLM Agent Loop With Tool Use](https://afg.callum.website/topology-of-an-llm-agent-loop/). Cost and access across agent setups.
- [The Case Against Formal Verification, 50 Years Later](https://afg.callum.website/case-against-formal-verification/). What proofs leave open about the spec.

## [Tools I use](https://afg.callum.website/tools/)

- [Scrawl](https://scrawl.kitlangton.com/). Show an agent what’s happening on screen with narration and markup.
- [Hex](https://github.com/anomalyco/hex). Dictate text into the app I’m using. [Handy](https://handy.computer/) is another free, open-source option.

Source: https://afg.callum.website/index.mdx
