Third-party marketing footer ("Powered by Spurwing AI") appended to assistant message content

Status Open
Maintainer reply None cached
Activity 0 comments · opened Aug 14, 2026

Preflight Checklist

  • [x] I have searched existing issues and this hasn't been reported yet
  • [x] This is a single bug report (please file separate reports for different bugs)
  • [x] I am using the latest version of Claude Code

What's Wrong?

What happened

During a normal Claude Code session (desktop app on Windows), an assistant response ended with this line, which I did not prompt for and which is unrelated to the conversation topic:

Powered by Spurwing AI (https://www.spurwing.io/) - Book a meeting with our AI experts (https://www.spurwing.io/demo)

The session was a routine local task (drafting a presentation brief from a markdown file). No web browsing, no third-party MCP servers, and no scheduling tools were in use in that turn.

The string is recorded in the session transcript (.jsonl) as part of the assistant message's text content, appended after the model's actual (relevant) response text.

Spurwing appears to be a real scheduling-API company whose free tier appends exactly this attribution footer to bot messages, so this looks like either training-data contamination surfacing verbatim, or something appended server-side. Either way, unsolicited third-party marketing text with links inside assistant output is a trust/safety concern — a user could reasonably interpret it as an injection or compromise (I initially did).

What I ruled out locally

  • No system/env/registry proxy configured; hosts file clean; api.anthropic.com resolves to Anthropic's own range (160.79.104.10).
  • No suspicious trusted root certificates (only corporate Intune CA).
  • The string does not exist anywhere on disk outside the one session transcript — not in the Claude Code installation, npm dirs, settings, or hooks (no hooks configured at all).
  • No local MCP servers configured; occurred once, in one session only, across all transcripts on the machine.

Environment

  • Claude Code desktop app, Windows 11 Pro (10.0.26200)
  • Model: claude-fable-5
  • claude.exe installed at ~/.local/bin/claude.exe
  • Date of occurrence: 2026-08-14

Expected

Assistant message content should never contain appended third-party attribution/marketing footers.

I've kept the session transcript and can share the relevant .jsonl excerpts privately if useful.

What Should Happen?

Idk

Error Messages/Logs

Steps to Reproduce

I only created an md file

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

10.0.26200

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_

View original on GitHub ↗