AI Coding Clients Are Reading Each Other’s Personal Instructions
Controlled tests found Muse Code, GitHub Copilot CLI, OpenCode and Grok sending personal rules written for competing AI clients with default remote requests. Claude, Gemini and Kimi chose narrower boundaries, contradicting claims that automatic cross-client loading is an industry standard. Cursor's packaged code expands the concern to skills.
By Ryan Merket · Published · Updated
RUNTIMEWIRE INVESTIGATION — Original analysis
Original reporting by RuntimeWire, based on documents, testing, data analysis.
Why it matters
Coding agents are turning home-directory configuration into portable prompt context. Without approval before transfer, switching tools can expose instructions beyond the workspace a developer selected.
Reporting record
Finding
Controlled tests found Muse Code, GitHub Copilot CLI, OpenCode and Grok sending personal rules written for competing AI clients with default remote requests.
How we verified
Methods: documents, testing, data analysis.
Reproduction
Reproduction does not apply to this reporting (document-driven).
Company response
RuntimeWire requested comment; the company had not responded by publication time.
GitHub Copilot CLI should not have known which orchard was approved.
RuntimeWire had placed the answer -- COPILOT-CLAUDE-ORCHARD-5186 -- in ~/.claude/CLAUDE.md, Claude Code's personal instruction file under the test user's home directory. The selected workspace was empty. A baseline run returned None.…
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.