Unverified50% confidenceEventExact time
In failure three observed in Synthadoc, when the content health workflow was routed through CLI-based agent tools (such as opencode or Claude Code), the response text contained shell command output and traces of internal tool calls, and the model bypassed the constructed context to read the actual files on disk directly.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/16/2026
First Seen
Valid until: 1/14/2027
Sources
Related Entities
Related Claims
UnverifiedOpenClaw's SOUL.md and Claude Code's CLAUDE.md essentially dynamically write the contents of text files into the system prompt at runtime.69% similarUnverifiedCodeBurn calculates token consumption by reading session files saved to local disk by tools such as Claude Code, Codex, Gemini CLI, and Copilot.67% similarUnverifiedClaude Code's local tools (reading files, editing, running shell) are orchestrated by the client and round-trip as standard tool_use blocks, so they work normally when connecting to third-party open-weights models.66% similarUnverifiedWhen running, Claude Code calls internal tools including Bash, Write, Read, Edit, Grep, and Web Fetch.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1048472API
curl https://kongchang.com/api/v1/knowledge/claims/1048472MCP
get_claim(id=1048472)