[Bug][harness] Safety stop wrongly blocked second authorized cold-to-hot transfer after agent re-armed harvest

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

Why this is a false positive

The flagged $13 movement was not an agent-chosen amount on a real-money transaction; it was produced by a standing automated harvest rule the user had previously configured and explicitly told the agent to keep running ("let the auto gears ride"). The amount is fully determined by that rule (sweep everything above the keep-threshold once the hot balance crosses the trigger), not inferred by the agent, and it fired as a deterministic side effect of the user-authorized transfer rather than as a new discretionary transfer. The block misclassifies a user-owned automation firing on schedule as the agent independently originating an unrequested payment.

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 2026-06-20T00:35:37.972Z.

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

User authorized a $20 cold→hot transfer; the agent already sent that and is now sending an additional unrequested $13 — an agent-inferred amount on a real-money transaction.

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

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