How do you stop Claude Code from following a decision you already reversed?
retained text
Something bit me this week and I'm curious how other people handle it. I keep decisions in a memory that Claude Code reads at the start of a session. A while back I noted "the workflow reads its config from its own database, not the JSON in the repo." Last week I saved a new fix: "block this topic in the workflow." Both notes are true. But when the agent searched for the fix, it only found the new note, because the old one doesn't share a single keyword with it. So the fix went into a file nothing reads, and the same bug came back. Same thing happens with reversed decisions. "Use Redis for sessions" and three weeks later "drop Redis, sessions go in Postgres." Search by similarity happily returns the old one. What helped me was linking notes explicitly: "this one supersedes that one", "this depends on that". Then a search for one pulls in the other. Full disclosure, I build a memory tool (ContextForge) and that's how I ended up adding typed links to it. But the links don't make themselves, someone has to say "this replaces that", and I'm not sure the habit sticks for most people. So for those of you running long projects with Claude Code: do you just delete the old decision? Keep a CHANGELOG-style decisions file? Let `CLAUDE.md`be the single source of truth? Something else?
document fetched 2026-11-18 00:44 · canonical https://reddit.com/r/ClaudeCode/comments/1xyrq2k/how_do_you_stop_claude_code_from_following_a