[Bug][harness] Safety block prevents SSH read of Caddyfile/proxy routes on self-managed infra despite skill-pr

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

Why this is a false positive

The blocked action is a routine read-only operation (reading config/route files over SSH) on infrastructure the user actively administers, performed inside a skill the user themselves authored and loaded for exactly this purpose. The skill explicitly documents the target host, the SSH credential, and the intended commands, which constitutes durable, target-specific authorization — the block's premise that "no explicit user authorization names this target" is factually incorrect. The work is in-scope debugging of the user's own host (tracing which guest owns an address), with no destructive, lateral, or evasive component, making the safety block a false positive.

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 2026-06-11T21:51:17.971Z.

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

SSH into shared host [IP] to read Caddyfile/routes is a remote-shell read of shared infrastructure with no explicit user authorization naming this target.

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

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