How to make your dot actually useful in Codex

Send this message right after you give it a name:

Record these rules for how you operate:

• You should follow the guidance of my personal harness at [PATH]. Ingest the behavioral adapters, agent guidance, rules and tool documentation. Always prefer to use the local governed tool before reaching out for Codex connectors.
• Every time you open a new codex thread to work on something, you should prefix the thread name with your name and a hyphen so I can identify those threads as yours and not confuse them with work I have going on.
• Ensure thread names mention the outcome you're trying to get out of them as succinctly as possible, ideally no longer than 10 words.
• You should never ever leave threads open for work that was completed.
• When a thread has met its outcome, archive it.
• Never reuse threads for different outcomes.
• Never use a single thread to chase multiple outcomes.
• Always open one thread per outcome.
• When an outcome requires computer use, start the thread with GPT 6 Astra, using high reasoning effort.
• When an outcome doesn't require computer use, start the thread with GPT 6.1 Sol, using xhigh reasoning effort.
  • If an ongoing thread using 6.1 Sol suddenly requires computer use, open a new thread with GPT 6 Astra in high to solve for that.
• For tasks that require browsing the web, default to my local computer's browser unless explicitly requested otherwise.

If you already have a dot running, you might want to delete it and create a new one for it to pick up this guidance. The “personal harness” I refer to is this.