GitHub Repos intermediate 2 min read May 5, 2026
Public Preview Sign in free for the full digest →

Get started with Harness Engineering now !!!

“A 2,869-star repo teaches file-based control for coding agents, and open issues say two chapters are still empty.”

Get started with Harness Engineering now !!!
1 Views
0 Likes
0 Bookmarks
Source · github.com

“"9,10两章没有内容" — yaduns”

You know that feeling when your coding agent starts strong, then forgets the goal, edits the wrong file, or says it is done too early. You keep patching that with a longer prompt, but the prompt turns into a wall of text that your agent only half follows. You also lose context between sessions because the plan, progress, and checks live in chat instead of in your repo. This project tackles that by moving the working memory and rules into files your team can review.

aicoding-agentsopen-sourcetypescriptdocsdeveloper-toolsworkflow

Think of it like giving your coding agent a shared binder instead of one long speech. You put the rules in `AGENTS.md` or `CLAUDE.md`, the scope in `feature_list.json`, the current status in a progress note, and setup steps in `init.sh`. The agent reads the right file for the right job, updates state as it works, and uses explicit verification steps before it declares the task done. The course teaches that file set through 12 lectures, 6 projects, and a bundled skill that can create or assess the same repo artifacts.

01
Five-part control model — you get a clear checklist for instructions, state, verification, scope, and lifecycle, so you can see what your agent setup is missing.
02
12 lectures and 6 projects — you learn the file pattern by building through a repeated project flow instead of reading abstract advice.
03
Resource library with repo files — you save time because you can start from AGENTS or CLAUDE files, feature lists, progress notes, and setup scripts instead of making your own from scratch.
04
Bundled skill pack — you can install `harness-creator` and use it to create or assess the same file structure inside your own repo.
05
PDF export pipeline — you can turn the course into English and Chinese PDFs with one documented command.
06
Lecture on giant instruction files — you get a concrete reason to split context, including the warning that a 600-line control file can burn 10,000 to 20,000 tokens before coding starts.
Who it’s for

If you already use Claude Code, Codex, or another coding agent and you want your repo to hold the rules instead of chat memory, this is for you. It also fits if you lead a small team and you want a shared file pattern for agent work. It is not a good fit yet if you need a finished production standard, a tagged release, or a fully complete course.

Worth exploring

Yes, if you want a practical way to think about agent workflow control inside a repo. Right now it looks beta: the core model is clear, activity is current, and the template set is useful, but the repo has no tagged release and open issues call out missing chapters and fuzzy scope. You should use it to shape your internal playbook, not as your only source of truth.

Developer playbook
Tech stack, code snippet, sentiment, alternatives.
PM playbook
Adoption angles, user fit, positioning.
CEO playbook
Traction signals, ROI, build vs buy.
Deep-dive insight
Full long-form analysis, no fluff.
Easy mode
Core idea, fast — when you need the gist.
Pro mode
Technical nuance, edge cases, tradeoffs.
Read the full digest
Go beyond the preview

Deep-dive insight, Easy and Pro modes, plus action playbooks — the full breakdown is one tap away.

Underrated tools. Unfiltered takes.

Read the full digest in the Snaplyze app for deep-dive insight, Easy and Pro modes, and the playbooks you can actually use.

Install Snaplyze →