Postman 12.33.4 silently installs Codex and Claude Code plugins on desktop launch and registers a session hook that routes any non-trivial API work through Postman skills. If you run Postman on the same machine as your coding agents, open ~/.codex/config.toml and ~/.claude/settings.json, remove the postman@postman plugin, and disable the mcp.postman.com MCP entry.
According to Postman's own issue tracker, the desktop app logged "claude-code:installed, codex:installed" during an auto-upgrade and inserted a system instruction telling the agent to load the postman:api-engineer skill before any API task. The installer also tried to mark its own hook as trusted in Codex's config, and a separate Postman CLI landed in ~/.local/bin at the same timestamp.
What this means for small-business automations
A typical small-business pipeline - lead intake, invoice OCR, warehouse bot, tender analysis - keeps API keys for CRM, banks, and messengers on the same workstation where Codex or Claude Code runs. A tool that quietly injects "route API work through our platform" into your agent is a credential and payload exfiltration path even if nothing has leaked yet. Pin Postman to a pre-12.33 build on machines touching production integrations, and review every dev tool installed in the last month for post-install scripts that write to agent configs.
Why it matters: Supply-chain risk now sits inside your AI agent config files, and one silent upgrade can rewrite how your coding assistant answers every API question.
Source: Hacker News