1:1 Notes Generator

Read a 1:1 meeting transcript from a file path and produce clean, structured notes ready to paste as a new entry at the top of that person's page in the "Lindsey 1:1s" Confluence space (L1).

Step 1: Read the Transcript

Read the file at the path provided by the user. Accept any format: plain text, markdown, auto-generated Zoom/Meet transcript, or rough notes. If the folder holds two transcript exports, use the later, larger one.

Step 2: Extract the Date

Look for a date in the transcript (in the filename, header, or body). If no date is found, use today's date.

Step 3: Generate the Notes

Produce output in exactly this format:

[DD Mon YYYY]

### TL;DR

[2-4 sentence high-level summary of what was discussed and any notable outcomes or themes.]

### [Topic Name]

* [Key point or note]
* [Key point or note]
  * [Supporting detail or specific example]

### [Topic Name]

* [Key point or note]

Nothing goes above the date and nothing follows the last topic: no title or participant heading, no ## Discussion Topics wrapper, no Action Items section. The first 1:1 (17 Aug 2026) had a trailing Action Items section; every entry since dropped it.

Formatting Rules

• Date: DD Mon YYYY (e.g. 05 Nov 2026) on its own line as the first line.
• TL;DR: 2-4 sentences max. Capture the overall tone and any big decisions or themes. Prose only, no bullets.
• Topics: Group related discussion points into named sections. Derive topic names from the conversation, don't use generic names like "Topic 1". Published entries run 4-6 topics; go shorter for short meetings, never pad.
• Notes under topics: Concise bullets. Capture decisions, concerns, context, and notable quotes. Strip filler and crosstalk. Sub-bullets are fine for lists of specifics.
• Actions: Record each action inline, in the topic where it arose, as prose: Lindsey will share the board link or Lindsey to send Christina small tickets. If the owner isn't clear, phrase it without one.
• Attribution: A Name: prefix marks whose view or statement a bullet is (* Lindsey: Christina is doing great). Don't use it for actions.
• Emphasis: No bold in the body. Bullets are plain text.
• Tone: Professional but not stiff. Match the register of the conversation.

Gotchas

• Auto-generated transcripts often have speaker labels like "Speaker 1:" or names. Use them to attribute actions and statements correctly.
• Zoom transcripts include timestamps. Ignore them for content, but use any date metadata for the note header.
• Small talk: skip opening pleasantries and closing remarks. A longer side conversation that reveals interests or working style can stay as a short topic at your judgment; it is occasional, not the default.
• If the transcript is very short (under ~5 exchanges), the TL;DR may be sufficient with one or two topics.
• Do not invent content. Only summarize what's actually in the transcript. If something is unclear, note it as "[unclear]" rather than guessing.

Output

Write the notes to artifacts/1on1-[YYYY-MM-DD]-[participant-name-if-detectable].md (ISO date in the filename, DD Mon YYYY only inside the note) and render:

bin/render artifacts/1on1-[YYYY-MM-DD]-[participant-name-if-detectable].md

If no participant name is detectable from the transcript, use artifacts/1on1-[YYYY-MM-DD].md.

Also print the notes directly to the conversation so the user can see them immediately.

Do not post to Confluence. The user pastes the entry into the page themselves.