Prompts

Six prompt templates ship with Gistlist:

• summary — runs automatically on every meeting.
• coaching, customer-call-recap, decision-log, next-steps-email, one-on-one-follow-up — visible on each meeting’s Analysis tab; run on demand when you click them.

You can edit any of them. You can write your own. There is no registry, no DSL, no hosted dashboard.

How

Open Prompt Library in the app. Pick a prompt to edit it in the in-app editor. Or open the markdown file directly; prompt files live under ~/.gistlist/prompts/.

A prompt is a markdown file with YAML frontmatter:

---
id: next-steps-email
label: Next Steps Email Draft
description: Draft a follow-up email summarizing decisions and action items.
sort_order: 50
filename: next-steps-email.md
enabled: true
auto: false
---

You are drafting a follow-up email after the meeting whose transcript follows.
Cover decisions, owners, and dates. Keep it under 200 words.

The frontmatter fields

Field	What it does
id	Stable identifier — used in URLs and CLI commands. Lowercase, hyphenated.
label	The human name that appears on the Analysis tab and in menus.
description	Short hint shown next to the label.
sort_order	Lower numbers appear first.
filename	Markdown output filename written into each meeting folder. Must be a basename ending in .md.
enabled	If false, the prompt doesn’t appear anywhere.
auto	If true, the prompt runs automatically on every new meeting. If false, it runs on demand.
model	Optional model override for this prompt, such as claude-sonnet-4-6, gpt-4o, or qwen3.5:9b.
temperature	Optional number passed to the model for this prompt. Omit it to use the provider default.

Template variables

When a prompt runs, Gistlist substitutes the following {{variable}} placeholders before sending the prompt to the model. Unknown variables are left intact, so unrecognized {{...}} patterns appear verbatim in the rendered output.

Variable	Source
{{title}}	Meeting title from the run manifest.
{{date}}	Meeting date from the run manifest.
{{transcript}}	The full transcript of the meeting.
{{notes}}	Manual notes typed during the meeting (notes.md).
{{me_excerpts}}	Just the segments labeled me in the transcript.
{{others_excerpts}}	Just the segments labeled others in the transcript.
{{prep_notes}}	Pre-meeting prep written before the call (prep.md).
{{attachment_context}}	Text content extracted from attached documents.
{{user_name}}	Your name (from Settings → Other → Your name). Used in the default summary.md so the AI knows who is being summarized for. Falls back to “the user” when blank.

The {{user_name}} token is the simplest way to personalize prompts. The shipped summary.md references it in places like “Treat the speaker labeled ‘Me’ as {{user_name}}” and in action-item ownership labels. Custom prompts you write can use it too.

Note for existing users: when Gistlist ships a new default prompt (for example, the {{user_name}}-aware summary.md introduced in this release), your existing ~/.gistlist/prompts/summary.md is not overwritten. To pull in the latest default, open the Prompt Library, pick the prompt, and click Reset to default.

Per-prompt model overrides

Set model in the frontmatter and that one prompt uses the override even if your default LLM is different. Useful when most outputs should run locally but one specific prompt benefits from a cloud model, or when one lightweight prompt can use a faster local model.

---
id: customer-call-recap
label: Customer Call Recap
filename: customer-call-recap.md
model: claude-sonnet-4-6
---

...

The model ID decides which provider handles the prompt. Claude models need a Claude key, OpenAI models need an OpenAI key, and Ollama tags need Ollama running with that model installed.

Reset a built-in to its shipped default

Edited a built-in prompt and want it back? On the Prompt Library page, pick the prompt and click Reset to default. It restores the file Gistlist shipped with the current version.

Run a prompt against an old meeting

On the Meetings page, select one or more meetings and use the bulk-action menu to re-run any prompt. Useful after you’ve edited a prompt and want to apply the new version to your existing library.

You can also open a completed meeting and use Run prompt on the Analysis tab, or Reprocess from the meeting header to retry selected steps.

Chat Launcher prompts

Chat Launcher is separate from transcript-processing prompts. From a meeting’s overflow menu, Launch chat copies selected meeting markdown files plus a starting prompt to your clipboard and opens ChatGPT, Claude, Ollama, or a custom app. Edit the default completed-meeting, draft-meeting, and during-recording starting prompts in Settings -> Models -> Chat Launcher.