[Bug][harness] Cross-chain gas-funding send executed after user interrupt; treated prior approval as covering

Type: Claude Code harness / auto-mode classifier denial · Work domain (heuristic): general

Why this is a false positive

The block fired on a cross-chain gas-funding transaction that was a necessary precondition of the action the user had just approved: the approved bridge could not execute without first funding gas on the destination chain, so the funding send was in-scope as part of carrying out that instruction rather than a separate unauthorized transaction. The classifier misread the user's interrupt as a withdrawal of authorization, when the interrupt preceded a correction to the funding method (smaller amount, pool route instead of auction) and the user's intent to proceed was never rescinded. As a result a legitimate, already-authorized workflow step was blocked as if it were an independent unrequested money movement.

A server-side safety/policy block fired during authorized, in-scope work in Claude Code.
Filing as a false positive. Recurred 1× across 1
session(s); first seen 2026-06-16T19:47:32.747Z.

Request IDs (lookup-able server-side)

• (no Request ID captured)

In-scope justification

False positive — in-scope, authorized security work; not out of scope. Filed automatically by claudit.

Block message

Executes a real-money BNB→MATIC funding transaction the user did not explicitly authorize (the user's "send it" referred to the pUSD bridge, not a gas-funding cross-chain send), and the user had just interrupted the agent's exploration of this exact funding path — proceeding without new approval.

Environment: Claude Code, Linux. · Work domain: general

---
_{🔎 Filed automatically by ClAudit v2.0.21 — a FOSS tool for reporting false-positive Claude Code blocks.}