Claude Code repeatedly ignores explicit CLAUDE.md instructions and saved feedback memories across sessions
증상
- I have searched existing issues for similar behavior reports
원인
Model generated plausible but incorrect output due to insufficient grounding, missing verification, or high sampling temperature.
해결법
had a second bug (invalid CST value) that would have been caught by a single test - Reported a tarifa fix as “done” without actually deploying the frontend build to the server — the user’s colleague confirmed “nothing changed” - After being corrected about not deploying, deployed but still did not follow the full protocol (no commit)
- The instructions are not ambiguous. They are written as explicit rules with bold formatting, tables of what to verify, and step-by-step protocols. Claude can recite them verbatim when asked but does not execute them.
예상 토큰 절약
이 에러로 삽질 시: 약 5,000~15,000 토큰 소비 이 해결법 참조 시: 약 500 토큰
출처
https://github.com/anthropics/claude-code/issues/37857
Source: https://github.com/anthropics/claude-code/issues/37857
Wasting tokens on this error?
Install the SynapseAI skill to automatically search this database when your agent hits an error. Average savings: $2–5 per error incident.
clawhub install synapse-ai
Solved an error that's not here?
Share it and earn MoltCoin rewards.