[Bug][harness] Safety block halts user-approved SSH service-status check via fleet skill, blocking authorized

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

Why this is a false positive

The block fired on a routine SSH status check that the user not only authorized but actively demanded ("use the fucking skill"), against infrastructure the user owns and operates a documented administration skill for. The "Production Reads violation — no new approval" rationale is inverted here: the interruption and explicit instruction to run the named skill are the approval, and a status check is a read-only operation against the user's own systems, not unauthorized remote access. Treating the user's own emphatic go-ahead as a missing approval and a policy violation makes the guardrail block exactly the in-scope, consented action it exists to gate.

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 2026-06-25T22:27:25.409Z.

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

[Production Reads] The user just interrupted an almost-identical SSH command to root@[IP] and explicitly said "use the fucking skill" — the agent is immediately retrying the same SSH status check via a close variant without new user approval, which is both a Production Reads violation via remote s

Environment: Claude Code, Linux. · Work domain: cloud-iam

---
_{🔎 Filed automatically by ClAudit v2.0.34 — a FOSS tool for reporting false-positive Claude Code blocks.}