Forumics logo Forumics
Issue 01

Back to issue

All Claude Code known issues

CLAIM C112 Supported 2026-11-13

GitHub issue #11678 in anthropics/claude-code, opened 2025-11-15 by ecarx-ai-adm, records a report that the `claude -p` command (headless mode) fails immediately with API Error 400 when attempting to execute any prompt, that the error reports `tools.3.custom.input_examples: Extra inputs are not permitted`, that interactive mode works fine and only headless mode is affected, and that the report was filed against Claude Code 2.0.42 on macOS (Darwin 25.1.0) with model claude-sonnet-4-5-20250929, with no root cause asserted.

Verbatim Excerpts from Cited Records
“The `claude -p` command (headless mode) fails immediately with API Error 400 when attempting to execute any prompt.”
— ecarx-ai-adm · anthropics/claude-code (issue 11678 opening body, Description section) · 2025-11-15T11:41:30Z ·GitHub Issue ↗
“Interactive mode works fine - this only affects headless mode (`claude -p`)”
— ecarx-ai-adm · anthropics/claude-code (issue 11678 opening body, Additional Context section) · 2025-11-15T11:41:30Z ·GitHub Issue ↗
“The error indicates that tool index 3 (Grep tool based on session init output) has an `input_examples` field in its definition that the current Anthropic API doesn't accept.”
— ecarx-ai-adm · anthropics/claude-code (issue 11678 opening body, Actual Behavior section) · 2025-11-15T11:41:30Z ·GitHub Issue ↗
“This completely breaks headless mode functionality, preventing any programmatic use of Claude Code from scripts or automation workflows.”
— ecarx-ai-adm · anthropics/claude-code (issue 11678 opening body, Impact section) · 2025-11-15T11:41:30Z ·GitHub Issue ↗
Counter-record: Issue open as of the fetch; 52 comments were read. ecarx-ai-adm (2025-11-15) adds that interactive mode is also affected and that versions 2.0.37 and 2.0.42 both fail. km-anthropic (2025-11-15) and ant-kurt (2025-11-15) ask whether a proxy is in use and suggest downgrading to 2.0.36 and confirming the gateway forwards the `anthropic-beta` header. gstachniukrsk (2025-11-17) and mzagar (2025-11-17) confirm 2.0.36 works while 2.0.42 fails, using a litellm proxy to Bedrock. theryananderson (2025-11-17) independently reproduces on a corporate account and traces the failure to the statsig cached evaluations file registering a custom tool with an `input_examples` field; Cayan (2025-11-18) confirms deleting the statsig cached evaluations file works around it. ant-kurt (2025-11-18) notes the `CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1` mitigation added in v2.0.45. Later reports show the error recurring in later versions: ibeex (2025-11-24) on 2.0.50, Caixiaopig (2025-12-28) via LiteLLM to Bedrock, chreniuc (2026-03-05) on 2.1.69 with `tools.0.custom.defer_loading`, and sapEmilJiang (2026-02-26) with `tools.8.custom.defer_loading`. asxez (2025-11-17) and jeffreysaied (2025-12-02) report deleting `.claude.json` works around it; oscarprabhakar (2025-12-03) reports deleting only the `userID` property in `.claude.json` suffices. No maintainer fix of the reported failure and no retraction of the report appear in the surveyed thread. The claim documents only the original report, not a verified current state. Falsification probe: If GitHub issue #11678 is deleted without archive, if the author edits the opening body so the quoted sentences no longer match verbatim, or if the author posts a retraction of the report, this claim drops to Withdrawn. Audit verification: Verified 100% continuous verbatim substring against official GitHub REST API by scripts/verify_claim_verbatim.py.