Skip to content

Conversation

@groovecoder
Copy link
Member

This PR fixes #.

How to test:

  • l10n changes have been submitted to the l10n repository, if any.
  • I've added a unit test to test for potential regressions of this bug.
  • I've added or updated relevant docs in the docs/ directory.
  • All UI revisions follow the coding standards, and use Protocol / Nebula colors where applicable (see /frontend/src/styles/colors.scss).
  • Commits in this PR are minimal and have descriptive commit messages.

Copy link
Member

@jwhitlock jwhitlock left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks @groovecoder. I feel like this is premature, and I'd prefer that you and other AI users collab outside the repo before committing. I am not an AI user, so I've done no testing of this file.

I suspect 500 lines is way too much, and will only work well with frontier models with large windows. I think 50 would be more appropriate.

More uninformed comments follow.

Copy link
Collaborator

@joeherm joeherm left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks good to me- I think it's worth having a discussion as a team before we merge it. I've played around with claude using the agents.md in the branch and think that it's giving good answers.

@groovecoder groovecoder self-assigned this Dec 29, 2025
@groovecoder groovecoder force-pushed the start-agents branch 12 times, most recently from 55e6d45 to 4debb9c Compare January 8, 2026 12:35
@groovecoder
Copy link
Member Author

Okay I did a significant refactor here.

  • I reduced the total content of instructions from around 3K lines to 769 lines. I also split those over a number of even smaller files (all about 100 lines or less) so agents can load only the instructions they need. And the smaller instructions should fit better into smaller local models' context windows.
  • I also added AGENTS.md to be more compatible with more agents, including opencode.
  • I removed the instructions about ops completely - I agree we don't want any agents performing any ops.
  • I removed almost all of the code examples and file listings instructions, as those can go out-of-date quickly, and agents seem good at exploring the codebase without using many tokens or time.
  • I removed a number of "common tasks" that were not common, as those are likely to pollute the context window.

Copy link
Collaborator

@Vinnl Vinnl left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for splitting it up, made it a bit more manageable for me. Gave some input on the frontend and testing piece.

Copy link
Collaborator

@joeherm joeherm left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks great- thanks for splitting it up.

Copy link
Member

@jwhitlock jwhitlock left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The markdown paths seem sloppy, but maybe this is standard for LLM stuff, so I stopped after a while. Don't wait for my approval.


The backend is organized into Django apps.

Check for an `.agents.md` file in an app directory for detailed guidance for that app.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

looks like it is called agents.md, not .agents.md

Comment on lines 116 to 119
- [privaterelay/.agents.md](privaterelay/.agents.md) - Core Django, settings, middleware, management commands
- [api/.agents.md](api/.agents.md) - REST API, authentication, serializers
- [emails/.agents.md](emails/.agents.md) - Email masking, AWS integration, metrics
- [phones/.agents.md](phones/.agents.md) - Phone masking, Twilio integration
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Files are agents.md (like privaterelay/agents.md), not .agents.md (like privaterelay/.agents.md). File is one folder up from here. So, should be something like:

[privaterelay/.agents.md](../privaterelay/agents.md)

Comment on lines 123 to 125
- [.agents.md](.agents.md) - Project overview and global principles
- [.agents/agents.frontend.md](.agents/agents.frontend.md) - Frontend development
- [.agents/agents.testing.md](.agents/agents.testing.md) - Testing guidance
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

paths are wrong here too

Comment on lines +129 to +130
- `docs/` - Comprehensive architecture documentation
- `README.md` - Full setup instructions
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

path appear wrong here too


### Single Frontend Static HTML Export

The frontend is built **once** and deployed to **all environments** (dev, stage, prod). Environment-specific values **cannot be baked into the build**. See "Environment-specific configuration" in [agents.backend.md](agents.backend.m) for more details. Next.js generates static HTML served by Django/Whitenoise. No server-side rendering. Dynamic routes must be handled carefully.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

path is agents.backend.md, not agents.md

CLAUDE.md Outdated
@@ -0,0 +1,5 @@
# Agent Lightbeam - Project Context
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

q: what is Agent Lightbeam?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

whoops, different project. nice catch.

@EMMLynch EMMLynch self-requested a review January 9, 2026 16:06
Copy link
Collaborator

@EMMLynch EMMLynch left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@groovecoder, thanks for the effort here and splitting out the agent files. Looks good.

@groovecoder
Copy link
Member Author

The markdown paths seem sloppy, but maybe this is standard for LLM stuff, so I stopped after a while. Don't wait for my approval.

Nah the paths were sloppy because I tried to manually fix all the references after I split up and moved the files. When I told claude to do it, it did a much more thorough job cleaning them up. Updated.

@groovecoder groovecoder enabled auto-merge January 9, 2026 17:49
@groovecoder groovecoder added this pull request to the merge queue Jan 9, 2026
Merged via the queue into main with commit 8487a23 Jan 9, 2026
30 checks passed
@groovecoder groovecoder deleted the start-agents branch January 9, 2026 18:10
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

7 participants