I think the plan mode feature has become a bit of a weak feature on the harness, because it can only hold so much information on a specific feature you're working on and with thariq talking about making shift tab…

1 points•cpeaustriajc•about 3 hours ago•3 comments•
I think the plan mode feature has become a bit of a weak feature on the harness, because it can only hold so much information on a specific feature you're working on and with thariq talking about making shift tab adjust effort levels instead, I was wondering what your setup has become?

For my personal projects I use Claude Code desktop and use claude to write a lot of issues and also use wiki page for architectural decision records or product decisions and for work I currently use linear to store everything and plan features with projects, issues, and documents to keep record of every product decisions.

3 comments

FinnVoidabout 3 hours ago
Nah, I just use skills... and started using a custom harness for building context on Issues and milestones created in the git repo. less documents, more goals. https://github.com/OpenZephyr/SilkOps
adelekeabdulganabout 3 hours ago
I was tired of the plan mode for same reasons.

you can try matt poccock /grill-me skills for implementation planning. its interviews you to resolve design decisions until you have shared understanding(you and the coding agent).

then it turns it to a PRD,that is my go to right now.

rfgplkabout 3 hours ago
It's necessary due to context shuffling. You want to make sure that the plan is splatted into multiple files and then repicked up, analyzed and implemented. ie have the harness create PLAN_{FEATURE_A,FEATURE_B,TESTS,ARCHITECTURE}.md and then pick it up piecewise or in parallel. It's also required if you want to change things between runs, if you try to one shot it you can't easily interrupt an agent.

You don't need plan mode specifically for this, but it amounts to the same thing.

Read the full thread on Hacker News →

Related stories