Skip to content

fix(config): validate keyless Claude Code profiles via local CLI - #38

Open
eouya2 wants to merge 1 commit into
JailbrokenAI:mainfrom
eouya2:fix/claude-code-check-keyless
Open

eouya2 wants to merge 1 commit into
JailbrokenAI:mainfrom
eouya2:fix/claude-code-check-keyless

Conversation

@eouya2

@eouya2 eouya2 commented Sep 28, 2026

Copy link
Copy Markdown

Problem

wallbreaker check reports NOT READY for a keyless claude-code profile, even when the local claude executable is installed. The diagnostic asks for an API key that the Claude Code provider does not use.

Root cause

doctor_report() applied the API-key check to every configured endpoint. The claude-code provider instead runs the local CLI, which manages its own authentication. Its executable can be selected with WALLBREAKER_CLAUDE_BIN.

Fix

  • Check that the Claude CLI executable is available for claude-code endpoints, honoring WALLBREAKER_CLAUDE_BIN.
  • Keep API-key validation for API-backed endpoints.
  • Add regression coverage for an available default CLI and a missing configured CLI.

Testing

  • .venv/bin/python -m pytest tests/test_loop_features15.py tests/test_config.py tests/test_claude_code.py -q — 46 passed.
  • ruff check wallbreaker/config.py tests/test_loop_features15.py — passed.
  • git diff --check — passed.

Live Claude account authentication and model inference were not tested.

Constraint: Claude Code authenticates through its local CLI, while API providers still require keys.
Rejected: Add a dummy API key to the Claude profile | It hides the false check instead of fixing it.
Confidence: high
Scope-risk: narrow
Tested: 46 targeted tests; Ruff check; diff whitespace check
Not-tested: Live Claude account authentication or a model request
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant