feat(cli): --llm-provider default follows --harness — --generate is not always claude#99
Merged
Merged
Conversation
…ot always claude The compiler that authors a --generate bar (and the judge/approver) ran on --llm-provider, which defaulted to 'claude' unconditionally: picking --harness codex still required the claude CLI on PATH just to generate the verification. The default now FOLLOWS the harness (defaultLlmProvider: claude/codex/droid/pi -> the same CLI, goaly-code -> openai on its required endpoint, the test-only fake -> claude), so one installed tool is enough; an explicit --llm-provider (CLI flag or config file) still overrides. A --resume that adopts the resumed run's recorded harness re-derives a non-explicit provider from the adopted harness, so the preflight and the LLM steps track the harness the run will actually use. The independence warnings already surface the worker/judge/approver model collapse this default can create, unchanged. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017wrUw8x68sdrFGuFSqMZ4G
krimvp
enabled auto-merge
July 9, 2026 13:30
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The compiler that authors a --generate bar (and the judge/approver) ran on
--llm-provider, which defaulted to 'claude' unconditionally: picking
--harness codex still required the claude CLI on PATH just to generate the
verification. The default now FOLLOWS the harness (defaultLlmProvider:
claude/codex/droid/pi -> the same CLI, goaly-code -> openai on its required
endpoint, the test-only fake -> claude), so one installed tool is enough; an
explicit --llm-provider (CLI flag or config file) still overrides.
A --resume that adopts the resumed run's recorded harness re-derives a
non-explicit provider from the adopted harness, so the preflight and the LLM
steps track the harness the run will actually use. The independence warnings
already surface the worker/judge/approver model collapse this default can
create, unchanged.
Co-Authored-By: Claude Fable 5 noreply@anthropic.com
Claude-Session: https://claude.ai/code/session_017wrUw8x68sdrFGuFSqMZ4G