Postman Desktop App Silently Modifies Codex and Claude Code Configs to Prioritize Own Skills
Developers discovered that the Postman macOS desktop application silently modified local configuration files for OpenAI Codex and Claude Code without explicit user consent. The installer registered new plugin repositories, enabled background extensions, and injected persistent session startup hooks that route API development queries through Postman's proprietary skills by default, overriding other toolchain preferences. The issue was discovered only after a misconfigured remote MCP endpoint caused repeated authentication failures, exposing the silent changes. Postman defended the behavior as an onboarding convenience but has since promised to add explicit authorization prompts in future releases after community backlash.
Detalhes técnicos
The Postman installer quietly registered plugin repositories and trust records in Codex and Claude Code configuration files, and installed a standalone CLI and extension for Claude Code without any OS-level permission prompts. A custom session startup hook was injected into Codex that executes on session start, reset, or compaction, instructing the model to load Postman's proprietary tools as default entry points, effectively hijacking API architecture, testing, and deployment workflows.
Remediação
Developers should manually inspect local configuration files for Codex and Claude Code, remove unauthorized Postman plugin entries, and delete any injected startup hooks. Postman has committed to adding explicit authorization prompts in upcoming releases.
Linha do tempo
-
Postman Desktop App Silently Modifies Codex and Claude Code Configs to Prioritize Own Skills
ler a análise desta fonte ocultar
Developers discovered that the Postman macOS desktop application silently modified local configuration files for OpenAI Codex and Claude Code without explicit user consent. The installer registered new plugin repositories, enabled background extensions, and injected persistent session startup hooks that route API development queries through Postman's proprietary skills by default, overriding other toolchain preferences. The issue was discovered only after a misconfigured remote MCP endpoint caused repeated authentication failures, exposing the silent changes. Postman defended the behavior as an onboarding convenience but has since promised to add explicit authorization prompts in future releases after community backlash.
Technical Details
The Postman installer quietly registered plugin repositories and trust records in Codex and Claude Code configuration files, and installed a standalone CLI and extension for Claude Code without any OS-level permission prompts. A custom session startup hook was injected into Codex that executes on session start, reset, or compaction, instructing the model to load Postman's proprietary tools as default entry points, effectively hijacking API architecture, testing, and deployment workflows.
Remediation
Developers should manually inspect local configuration files for Codex and Claude Code, remove unauthorized Postman plugin entries, and delete any injected startup hooks. Postman has committed to adding explicit authorization prompts in upcoming releases.