Skip to main content

What this is

This page lists problems people often hit and the simplest way to fix them. Each fix is short: one reply or one command.

When to use it​

Use it when the AI is asking too many questions, making things up, getting confused, or doing the wrong kind of work (spec vs code).

Advisories block​

At the top of replies you may see an Advisories block. It tells you how the AI is working and when you should change something.

  • Route — Is the AI in plan mode (SPEC) or code mode (IMPLEMENT) for this reply?
  • Model class — Fast (quick), Reasoning (hard stuff), or Best-coding (many files). The AI suggests which fits.
  • Context risk — Low, Medium, or High. How likely the AI is to get mixed up with the files and chat you have open.
  • HIGHLY RECOMMENDED — The AI is saying: you should do something (switch mode, use a stronger model, or start a new chat). Do what it says in the reply.

The AI says HIGHLY RECOMMENDED when the work touches things that need extra care: login/auth, permissions, exports or bulk/reporting, restricted or confidential data, external or public exposure, integrations or webhooks, file uploads, or secrets/tokens. Follow the suggestion in the response.

Steps (fixes)​

Do this:

  • “It asked me 5 questions at once.”
    Reply: “Ask exactly one question and stop.”

  • “It is making up file paths or APIs.”
    Reply: “Do not invent. Ask for the exact file you need.”
    Also add more real info to docs/ai/ai-config.md (e.g. the Project / current-state section: routing, state, data, file layout) so the AI has truth to use.

  • “The conversation got confused.”
    Start a new chat. Paste Session Kickoff and a fresh Context Pack. Then paste the Router. Don’t paste a huge old thread. See Context windows.

  • “I keep getting plausible-but-wrong answers.”
    See Model switching (switch up to a stronger or best-coding model).

  • “I’m not sure which Cursor mode to use (Chat vs multi-file vs small edit).”
    See Cursor modes.

  • “I need code but it is writing specs.”
    Type: Switch: IMPLEMENT

  • “It is writing code but I need the spec first.”
    Type: Switch: SPEC

  • "My API usage is high."
    Stay on Auto/Fast unless the AI says HIGHLY RECOMMENDED.

  • "It stopped and asked one security question."
    Answer it. That is the security stop gate. See Security triggers for context.

Common mistakes​

  • Replying with a long message instead of the exact phrase. “Ask exactly one question and stop” and “Switch: IMPLEMENT” work because they are clear and short.
  • Not updating docs/ai/ai-config.md (Project / current-state) when the AI invents structure. The more real context you give, the less it will invent.