CodeMonument
codemonument.com
CodeMonument
@codemonument.com
Ohhhhh mega! 😱😍🙌
January 28, 2026 at 10:26 PM
Yooooooo Mega cool, directly integrated with Tab!
January 28, 2026 at 4:36 PM
Jooo, I looked for this some time ago and was confused that this is kinda complicated with pure rust 😄 thank youuuu!
January 28, 2026 at 4:35 PM
Interesting! I'll probably use that! What do you use zig for in this repo?
January 28, 2026 at 12:16 PM
Because, for example, I reference some explicit skills the agent should use or some workflow requirements like "always commit after adding a json ticket in folder x".

Human devs don't need this sort of information, they most know the skill already or do their workflow as they please.
January 26, 2026 at 6:18 PM
But the guidelines I give to agents are more technical than I'd give to humans
January 26, 2026 at 12:30 PM
Same here 😩
January 25, 2026 at 8:35 PM
Yes, after a slow start I'm also reading more about it.
If I want to selfhost my pds, would you still go with the bluesky pds offering or is there something better at the start of 2026 now?
January 25, 2026 at 12:55 PM
(I'm from Germany too, Leipzig currently, therefore the excitement :D)
January 25, 2026 at 12:54 PM
This is so cool! I'm really rooting for you that chatting flies 🙌
January 25, 2026 at 12:53 PM
It finds the beans skill automatically when I mention beans and loads the "beans prime " command.

Maybe this is a setup you like 👍
January 24, 2026 at 9:32 AM
That would be great!

Also: I made me a beans SKILL out of the --help of the cli and added the "beans prime" rule as well as my custom "bean creation rules" to it. That way I don't have to specify beans prime in each Agents.md in each of my projects, because
AGENTS.md
AGENTS.md is a simple, open format for guiding coding agents. Think of it as a README for agents.
Agents.md
January 24, 2026 at 9:32 AM
And especially by creating beans for committing, it needs to commit the beans too 😂
January 24, 2026 at 8:49 AM
I can start the loop for three projects at ones which i prepared earlier when i now: the agent can use my full 5 hour anthropic session window to get work done.

And when i come back, i only have to tell it to fix some bugs here and there and boom - finished!
January 23, 2026 at 9:42 PM
After that i tell opencode with oh-my-opencode ralph loop to:

grab one bean, work on it, then finish it or mark it as blocked and work until all beans are done or blocked (aka. need user intervention).

This is not fully autonomous yet, but it gets work done in huuuuge chunks!
January 23, 2026 at 9:42 PM
I think you may also be able to set some more security settings, like who is allowed to publish, but I did not use that yet, since I publish alone right now 😅😂
January 23, 2026 at 8:22 PM
You can look at the "provenance" deployment logic for jsr, I like it a lot:
You go there, say "create new package", then add your GitHub repo name + your username or so. Finished. GitHub builds from there are allowed to publish to jsr. Nice!
January 23, 2026 at 8:22 PM