Skip to content
esc

Suggested

    ↑↓ to move↵ to open⇧ ↵ to see all results

    Connect your AI

    Things to ask

    Here are some starter prompts to try once your AI is connected.

    Your AI reads your last week across email, calendar, files and chat, shows you what it plans to save, then shows you the difference your context makes.

    My Corridor Context brain is connected. Use the connectors you have for my email, calendar, files and chat, plus these tools: cold_start_context, personal_list, personal_search, personal_capture, personal_update, personal_log.
    1. Read: cold_start_context, then personal_list, so you know what my brain already holds. Then read the last 7 days from every source you can reach: email (including sent), calendar, files, chat. If a source isn't connected, skip it and tell me in one line. If my brain is empty, treat everything as new.
    2. Find what my brain is missing or has out of date:
    - how I write when I write well (greeting, length, sign-off, phrases I reuse, things I avoid), taken only from emails I actually sent
    - what I'm working on, with status and my role
    - the people I work with most, with role, company, and how we work together
    - decisions made and dates coming up
    Skip anything my brain already has unless it changed. Never save passwords, API keys, account numbers, or anything sensitive. Mention once that you skipped them.
    3. Show me the plan before saving: grouped, numbered, one line each, marked NEW or UPDATE, with the source (email subject, meeting title, or file name). If anything rests on thin evidence, say so. Wait for my go. I may drop items by number.
    4. When I say go: save each item where it fits best (new page with personal_capture, a fix with personal_update, anything else with personal_log). Then show me what you saved.
    5. Then build one inline visual that shows me why this matters. Use only what you just saved, with real numbers and real names from my brain. Never invent counts or examples. Show the same 3 requests I'd realistically make, answered two ways side by side: without my context, and with it. Pick requests where the context changes the answer: a follow-up email in my voice to a real person, "what do I have coming up", and a status check on a real project. Under each "with context" answer, list in one line what it knew and where it came from. Make the three requests switchable.
    6. End with the three biggest gaps as questions I can answer in one line each.

    Close to the prompt at the end of setup: your AI reads your brain and shows you three things it can help you finish, as cards you can start from.

    My Corridor Context brain is connected. Do steps 1 to 3 in one reply without asking. Use my people's and projects' names, not generic labels. Tools: cold_start_context, personal_list, personal_read, personal_search, personal_update, personal_capture.
    1. Read: cold_start_context, then personal_list. Skim the newest raw captures first, then read at most 5 people/ or projects/ pages, one source page per workflow. Rank workflows by immediacy, leverage, and a visible result today.
    2. Fix quietly on the pages you read: personal_update wrong or stale facts; mark passed dates with no recorded outcome as unresolved; personal_capture a missing person or project a workflow needs. Do not report fixes.
    3. Build one interactive visualization (HTML). Header: greet me the way communication-style.md says, or just my first name if it says nothing about greetings; one line on my role; three one-line bullets on what I am trying to get done this month and where the friction is. Then 3 numbered cards, each tied to a named project, person, or deadline: a bold title in my words; one quote from the source page, 15 words max; if a second page or record confirms or contradicts it, one line saying so, otherwise skip; the first move; time to first result; what gets saved; days left if a date is attached; a Start control that sends "Start N" to this chat. Outside the visualization write only: Reply with a number to start.
    4. When I reply with a number or Start N: re-read what you need, do move one now with my tools, then guide me through moves two and three. Finish by saving a playbook with personal_capture type playbook, named in my words: trigger, steps, sources, tools used, what to capture, and the shortest way to rerun it here.

    Your AI lays out your week and the one or two things coming up that matter most.

    My Corridor Context brain is connected. Do steps 1 to 3 in one reply without asking. Use real names, not generic labels. Tools: cold_start_context, personal_search, personal_read, meeting_search, meeting_details, company_search.
    1. Read: cold_start_context; my calendar for the next 7 days if you can see it; meeting_search for my meetings in the last 7 days; then the pages for the projects and people involved.
    2. Work out what I'm working on, who's involved, what's waiting on me (things I said I'd do, asks I haven't answered), and the one or two things coming up that matter most and why.
    3. Build one visualization (HTML) in light mode, with short sentences that are easy to read and understand: my week as a day-by-day strip with the meetings that matter; a "Waiting on you" list, each with who is waiting and since when; and one or two highlighted cards for what matters most, each with why it matters, what to prepare and the first move. If my brain was missing something you needed, end with up to three questions.
    4. When I answer, save my answers with personal_capture or personal_update.

    Join late and ask what you missed, or get the last meeting the notetaker attended.

    Check whether my Corridor Context notetaker is in a meeting with me right now. If it is, catch me up on what's been said so far and what's been decided. If not, use the most recent meeting it attended. Either way, list the action items with who owns each one, pull out anything I said I'd do, and draft a short follow-up email to the attendees.

    Next: see what your AI can do with each tool.

    View as Markdown