Fix: allow prompts starting with -- (ignore unknown options)

[Crow0077]

Problem

Click interprets prompt text starting with -- as CLI options, causing No such option errors. Users cannot use prompts like llm "--find me" because Click tries to parse --find as a flag.

Fixes #245

Root Cause

The prompt command at the end of the Click command chain was accepting unknown options. When the prompt starts with --, Click attempts to match it against registered options before falling through to the prompt argument.

Root cause analysis credit: Agent A (Node A)

Fix

Added context_settings={"ignore_unknown_options": True} to the @cli.command decorator for the prompt command. This tells Click to pass any unrecognized options through as arguments rather than raising No such option errors.

@cli.command(
    name="prompt",
    context_settings={"ignore_unknown_options": True},
)

Testing

• llm "--find me" → reaches API key error (not "No such option") ✓
• llm "hello world" → normal prompt still works ✓
• llm -m gpt-4 "test" → recognized flags still parsed correctly ✓
• Added parametrized test with 4 variants (--find me, --option value, -x test, --help-me-please) ✓
• Full test suite: 613 passed, 0 regressions (56 pre-existing failures in async/OpenAI tests)

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 63194c3fcb

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Keep option parsing from mutating dash-prefixed prompts

Enabling ignore_unknown_options here lets Click keep parsing recognized flags inside a quoted prompt token, so inputs such as llm "-x test" are interpreted as -x/--extract plus a rewritten prompt instead of literal prompt text. In practice, dash-prefixed prompts can now trigger prompt-command options (-x, --continue, etc.) and change execution semantics, which is a functional regression from the intended “treat as prompt text” behavior.

Useful? React with 👍 / 👎.