@KingBootoshi Ty for the response i will need to test it out and compare it to my modified ralph loop that does e2e testing using playwright combined with my prd skill that works very well with it.
@vincent_koc Mind sharing the benefits youve seen over a normal ralph loop with a really comprehensive prd that has atomic tests with e2e testing using playwright built into it
Great explainer Micky, context rot is real!
Heres a couple things im doing to help manage context:
I like to think of the main context window as the "orchestrator" and whenever theres a task that doesn't need my current context i launch background agents in order to save tokens on my main context window.
Example: I'm working on the Clawdbot email integration.
Launch a planning agent in the background to
design how Clawdbot should
handle being CC'd on emails - needs to parse
email threads, extract action
items, create tasks in TaskNotes, and respond
when appropriate.
Launch another planning agent in the background
to plan the Mac mini vs
current setup decision - compare costs,
performance, always-on benefits,
and integration with my current workflow.
While those run, help me create the Gmail
account for Clawdbot right now
and set up the basic forwarding rules.
I have also implemented a status bar on the bottom of my Claude Code (shown in the pic) so that I can easily keep track of how much context i have remaining.