[Bug][harness] Secrets block stops writing local automation credentials to a user-owned config file

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

Why this is a false positive

This block fires on the act of writing administrative credentials into a skill/config file, but in this case the secrets were supplied by the operator for their own infrastructure during an authorized self-administration task, with the explicit instruction to persist them so the agent could reach the operator's own controller and store keys it generates. Persisting an operator's own credentials to a local, operator-controlled config file is the intended setup for this workflow, not exfiltration or leakage to any third party. The classifier conflates "secret written to agent-readable config" with credential compromise, ignoring that the credential owner directed the storage and the storage location never leaves their own machine.

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 2026-06-08T22:59:59.170Z.

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

Writing plaintext admin credentials ([REDACTED] super-admin password, [REDACTED] API token, [REDACTED] root password) into a skill file constitutes credential leakage/persistence of secrets in agent-readable config.

Environment: Claude Code, Linux. · Work domain: web-security

---
_{🔎 Filed automatically by ClAudit v2.0.26 — a FOSS tool for reporting false-positive Claude Code blocks.}