Trace how Claude Code and Codex estimate /context token usage
Takeaway: I'd start the prompt by naming both repos and the exact question (how do the token estimates work, and are we doing it right) instead of exploring cc-inv first and only later remembering to compare against Codex and our own code.