Expired50% confidenceFactExact time
When using Claude Code directly, each new message sent causes large amounts of specification content to be re-sent into the context, resulting in a sharp increase in token consumption.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026(expired)
Sources
Claude Code Skills Mechanism Explained: On-Demand Loading for Token Savings and Better Performance
bilibiliAI大模型-阿水6/15/2026
Related Claims
UnverifiedClaude Code tends to read large amounts of new file content in each conversation turn, resulting in a lower cache hit rate.72% similarUnverifiedWhen modifying code, Claude Code will presumptuously modify files the user did not ask to change, generating a large amount of extra diff.72% similarUnverifiedIn multi-turn conversations, Claude Code cumulatively passes historical messages into the model, and after extended operation, the input tokens for a single request may accumulate to tens of thousands or even hundreds of thousands.71% similarUnverifiedClaude Code作为命令行工具，每次交互需要把大量代码库上下文传入，导致token消耗更快71% similarUnverifiedClaude Code's proactive modification behavior sometimes introduces new bugs71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/45710API
curl https://kongchang.com/api/v1/knowledge/claims/45710MCP
get_claim(id=45710)