Compact prompt should preserve historical context across multiple compactions
증상
When auto-compact fires, the default prompt focuses on recent work. After 2-3 compactions in the same session, all reasoning, decisions, and context from earlier compactions is completely lost. The model doesn’‘t even know there WAS a previous compaction.
원인
Agent session state was not persisted to durable storage, causing context to be lost on restart or session switch.
해결법
above) | | 6-8 | Acceptable | High-level road preserved, implementation details gone | | 9+ | Degraded | Even cumulative summaries get compressed, signal-to-noise drops |
Recommendation for docs: Switch sessions when the task is DONE. One session = one task/feature. Not one session forever, not a new session every 30 minutes. Without this guidance, users either restart too often (losing warm context - files read, patterns learned, user preferences) or never restart (context becomes soup after 10+ compactions).
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/33212
Source: https://github.com/anthropics/claude-code/issues/33212
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.