feat: use generated AI docs in local circuit prompt

[LYL-8bit]

Summary

• Switches the local circuit prompt from the older generated component props markdown to the auto-updating tscircuit AI docs super context at https://docs.tscircuit.com/ai.txt
• Adds explicit prompt guidance that the fetched docs are auto-generated and should be preferred over stale hand-written notes
• Keeps existing footprint/rules context intact

Validation

• ./node_modules/.bin/tsc --noEmit
• Verified https://docs.tscircuit.com/ai.txt returns HTTP 200 and the Repomix-generated docs payload

Closes #45

[LYL-8bit]

CI note: the type-check workflow is green. The Bun Test workflow is failing because the fork/PR run has no OPENAI_API_KEY available (generateRandomPrompts gets OpenAI 401, and the coder integration test returns no generated VFS source). I only changed the local prompt template docs source, and verified locally with ./node_modules/.bin/tsc --noEmit.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a85688371f

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]

Bound AI docs payload before injecting into system prompt

This change replaces a narrowly scoped component-props doc with ai.txt, then injects the entire fetched payload into the system prompt without any size cap or section filtering. ai.txt is a repomix-style aggregate of many docs files, so prompt size can grow to the point that benchmark runs either fail with context-length errors or lose most of the response budget to background context instead of the user task. Please truncate and/or select only the component/props sections before interpolation.

Useful? React with 👍 / 👎.