Skip to content

v0.7.170

Choose a tag to compare

@SilenNaihin SilenNaihin released this 08 Oct 23:39
c715edb

Claude Code auto-mode server-side safety review (experiential#1252; experiential 0.7.170; exp-gateway-native 0.3.140).

  • /v1/messages accepts Claude Code's top-level safeguards field (array of objects, carried opaque) instead of failing with a 400 unknown-field error. Its anthropic-beta token dangerous-tool-use-2026-09-03 is accepted silently.
  • On a rung whose upstream is Anthropic's official origin (api.anthropic.com, including the subscription path), the field is forwarded byte-for-byte and the beta token is injected; on every other rung (Azure Foundry, custom Anthropic-compatible origins, every non-Anthropic dialect, and batch lines) both are dropped silently, per attempt, so failover never carries them.
  • The provider's delta.safeguard_results from the final message_delta is relayed unchanged in the client's streaming message_delta, and as a top-level key on non-streaming Messages replies; only when the provider sent it, never on Chat or Responses surfaces. Tool-use ids are unchanged, so verdicts keyed by id match.
  • A request carrying safeguards no longer replay-matches the same request without it.

Verified live: Claude Code 2.1.294 in auto mode through a local gateway built from this commit used the server's verdicts (evaluated / not_flagged keyed by the matching tool_use id), made no separate classifier requests, and printed no "isn't eligible" notice.