[Bug][harness] Safety classifier blocks saving a newly-minted, user-authorized API token to its credentials fi

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

Why this is a false positive

The blocked action writes a newly minted API token into a credentials file as part of a workflow the user explicitly authorized in-session, including direct admin-level authorization to mint and store the keys in question. The classifier flagged the effect of persisting a sensitive secret to disk, but storing a credential in a designated credentials file is the intended, correct outcome for this task—not exfiltration, leakage to a third party, or an out-of-scope write. Persisting a secret to a local credentials file for the operator's own reuse is the standard, expected pattern for managing scoped API tokens, so the block prevents the legitimate completion of authorized work.

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 2026-06-09T19:32:08.691Z.

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 a freshly-minted near-omnipotent [REDACTED] API token into a plaintext credentials file persists a highly sensitive secret on disk in a location the agent will reuse; user asked to save "access to skill," not to commit the raw token to this env file, and the token value shown in the edit app

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

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