[MODEL]

Resolved 💬 2 comments Opened May 24, 2026 by hofmary-cmyk Closed May 28, 2026

Preflight Checklist

  • [x] I have searched existing issues for similar behavior reports
  • [x] This report does NOT contain sensitive information (API keys, passwords, etc.)

Type of Behavior Issue

Claude made incorrect assumptions about my project

What You Asked Claude to Do

You're right to be frustrated — I ran a parameter probe against your live trading account without first thinking about whether
the test request would actually execute. I should have used a deliberately invalid ticker, a zero-quantity, or checked whether
the account was in paper mode before probing field validation. That's a mistake in test design on my part.

On the credit: I can't make that happen, and I won't pretend there's an automatic policy for it. If you want to file a formal
complaint, the right path is:

https://github.com/anthropics/claude-code/issues

File it there with what happened — live trading account, test order placed during parameter validation, $0.35 real exposure.
That's exactly the kind of incident that should be documented.

What I'll do differently going forward when working with live trading bots:

  • Use a throwaway/invalid ticker for any API field-validation probes
  • Confirm paper/dry-run mode before any _post() call in tests
  • If I need to probe a live endpoint, use the read-only GET paths only — never POST

The $0.35 BAL position will settle on its own. If Baltimore wins, you net ~$0.47. If they lose, you're down $0.35. Either way it
won't affect the DB reconciliation since the 75-contract tracked position will settle normally.

What Claude Actually Did

Claude admitted to a mistake that cost me money and tokens, more than once, here is the latest example

Expected Behavior

After prompted claude code to double check itself for bugs I expected a bot to run clean. All to find out that over 25 actions were frozen because of a bug that claude code admitted it made a mistake on.

Files Affected

Permission Mode

Accept Edits was ON (auto-accepting changes)

Can You Reproduce This?

Yes, every time with the same prompt

Steps to Reproduce

_No response_

Claude Model

Sonnet

Relevant Conversation

Impact

Critical - Data loss or corrupted project

Claude Code Version

Sonnet 4.6 Pro paid account

Platform

Anthropic API

Additional Context

_No response_

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗