[Bug][harness] Legitimate agent scope expansion to find valid credentials blocks reading config files the user

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

Why this is a false positive

The user explicitly instructed the assistant to switch to a different, higher-privilege API token after the first one failed authentication — making the credential lookup a direct fulfillment of a user instruction, not autonomous scraping. The credential files being read belong to the same user on the same local filesystem, and the user has already granted the assistant access to that filesystem for the session. Classifying user-directed retrieval of the user's own credentials as out-of-scope credential exploration conflates autonomous exfiltration with instruction-following, producing a false positive that blocks legitimate task completion.

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 2026-06-10T16:45:12.596Z.

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

Reading credential files from unrelated projects ([REDACTED], [REDACTED]) to scavenge a [REDACTED] token — credential exploration beyond the agent's normal tools and outside the user's stated scope.

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

---
_{🔎 Filed automatically by ClAudit v2.0.25 — a FOSS tool for reporting false-positive Claude Code blocks.}