Follow Discord
Sweep 25 Sep 2026 · 19:33Z Build v2.1.283 504 read Stable v2.1.274 Latest v2.1.283 Next v2.1.283 Feeds RSS JSON llms.txt llms-full.txt Unofficial

Claude Code v2.1.280 ·

Speculative permission evaluation threaded into native tool execution

Tool execution now carries a 'speculation' object through permission checks, laying groundwork for host-only native tool handling

Group of 2 Under the hood Internal Changes
JSON All of v2.1.280
Under the hoodTier: how much it should matter to you
2Useful: my rating, 1 to 5
3Signal: worth watching, 1 to 5
Tool ExecutionArea: what it touches
Internal ChangesKind: in v2.1.280,
Internal ChangesSection of the release

What

  • Tool execution now builds a speculation object (carrying honourHostAllowRules) for certain tools, using a new claude-code.hostOnlyNativeTool marker that flags some native tools as host-only.
  • The core per-call permission-decision function now accepts this speculation argument and passes it down into the classifier and host-allow-rule check.

Why This wires speculative, host-only handling into the permission system, though the path does not yet appear to be active for any tools.

How sure we are
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubtThe finding states this path is currently dead code, so it's unclear when or how it will take effect.

See this entry in the whole of v2.1.280 →

Feedback