Learn concrete phrasing to tell a coding agent so it tests its own work automatically.
Adopt a pattern for storing project memory in files so agent sessions can restart cleanly.
Pick up house-style advice on writing plain, framework-light code that is easier for an agent to verify.
| anode1/agent-recipes | 00kaku/gallery-slider-block | 04amanrajj/netwatch | |
|---|---|---|---|
| Stars | 0 | — | 0 |
| Language | — | JavaScript | Rust |
| Last pushed | — | 2021-05-19 | — |
| Maintenance | — | Dormant | — |
| Setup difficulty | easy | easy | moderate |
| Complexity | 1/5 | 2/5 | 3/5 |
| Audience | developer | general | ops devops |
Figures from each repo's GitHub metadata at analysis time.
This repository is a collection of practical tips for working with AI coding agents, written as short recipes rather than a piece of software you install. Each recipe has a one line instruction you either say to the agent or a practice you follow yourself, followed by a short paragraph explaining the reasoning behind it. The core idea running through the recipes is that an agent only truly knows what is written down in the project's files. Whatever lives only in your head or in a chat conversation disappears the moment that session ends, so the author argues you should keep goals, state, and checks inside the files themselves rather than relying on memory. One major theme is testing. The author recommends giving every feature, including visual interface work, an automated test the agent can run and check on its own, without a person watching. For web pages this means using a headless browser to take screenshots and compare them programmatically, and for other kinds of software it means a normal automated test suite. Two small companion tools the author built, one for taking those screenshots and one for message passing between multiple agents, are linked as examples. Another recurring piece of advice is to start a new conversation with the agent once the current one gets long, rather than trying to keep working in an overloaded conversation history, and instead to store important facts in a small searchable memory tool so they can be looked up again cheaply later instead of being explained from scratch every time. The later sections lean more personal, describing the author's own preferences, such as writing plain code without heavy frameworks so an AI agent has less to understand and fewer ways to get things wrong, keeping web backends simple by returning results straight from a database query, and writing C code in a careful, restricted style similar to standards used in safety critical industries like aviation and medical devices. There is no programming language listed for the project itself, since it is really a written guide plus a handful of small example files and companion tool links rather than an application.
A written guide of short practical recipes and best practices for driving AI coding agents effectively.
No license information was found in the description provided.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.