A token-lean fork of an early ECC snapshot (~1,260 always-on tokens) #2651
OsamaA140
started this conversation in
Show and tell (with ECC)
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Thanks for open-sourcing ECC — I've learned a lot from it.
For my own daily use I maintain a compressed fork of an early snapshot (9 agents, 26 skills): same coverage and behavior, but rewritten for minimum context cost — roughly 70-80% fewer tokens per invocation and about 1,260 always-on tokens for the whole stack. Enforcement lives in zero-token hooks (secret-leak blocking before writes hit disk, destructive-command guard), with the guard rules unit-tested (100/100) and CI on Node 20/22.
Repo: https://github.com/OsamaA140/everything-claude-code-lean
MIT, full credit to you in the README and OPTIMIZATION.md (which documents exactly what was compressed and why). If any of the token-optimization approach is useful upstream, happy for any of it to be taken — and thanks again for the foundation.
All reactions