answer_input — Claude Code prompt: what it does, how to install
The answer_input Claude Code prompt builds an answer from meeting context. It can use spoken lines, calendar details, photos, and notes from earlier meetings.
MIT ★ 395 updated 2026-11-16
.claude/commands/answer_input.mdWhat answer_input does
The prompt gathers meeting context to answer a question typed on a phone. If no question was entered, it uses the lines from the 45 seconds before the tap along with details such as the meeting title, attendees, language, location, prior questions, and photos with captions or recognized text.
Earlier meeting notes can contribute discussion points, decisions, action items, open questions, and project status. When earlier minutes are unavailable, the context contains (none with minutes). In Claude Code, place the command file at .claude/commands/answer_input.md.
What it helps with
- answer a question using the current meeting context
- recall what was said just before a tap
- find decisions and action items in earlier meeting notes
Example prompts for the agent
- Answer this from the meeting: who will take the task?
- Summarize what was said just before I tapped.
- Which decisions and action items from earlier meetings relate to this project?
How to install answer_input
Manually: put the files in place
.claude/commands/answer_input.mdHow to use
Paste the text at the start of a chat or into project instructions
Questions about answer_input
How do I install answer_input in Claude Code?
Place the command file at .claude/commands/answer_input.md.
What does answer_input do?
It builds an answer from the current meeting details and can also draw on notes from earlier meetings.
What happens if I don't type a question?
The prompt uses the lines from the 45 seconds before the tap and the current meeting context.
Source text
answer_input.md · MIT
This meeting, being recorded now (times are minutes:seconds from its start):
<this_meeting>
{{with .Title}}Title: {{attr .}}
{{end}}{{with .Calendar}}{{with .Title}}Calendar title: {{attr .}}
{{end}}{{with .Attendees}}Calendar attendees: {{range $i, $a := .}}{{if $i}}, {{end}}{{attr $a}}{{end}}
{{end}}{{with .Location}}Calendar location: {{attr .}}
{{end}}{{end}}{{with .Language}}Language: {{attr .}}
{{end}}Said so far (the latest lines, oldest first):
{{range .Lines}}[{{.At}}] {{attr .Text}}
{{else}}(nothing yet)
{{end}}{{with .Asks}}Already answered in this meeting:
{{range .}}- Q: {{attr .Question}}
A: {{attr .Answer}}
{{end}}{{end}}{{with .Photos}}Photos taken in it (the owner's caption, the text recognized in it on the phone, a note on it):
{{range .}}[{{.At}}] photo {{.N}}{{with .Caption}} · caption: {{attr .}}{{end}}{{with .Text}} · text: {{attr .}}{{end}}{{with .Note}} · note: {{attr .}}{{end}}
{{end}}{{end}}</this_meeting>
{{if .EarlierAvailable}}Earlier meetings, by their minutes:
{{range .Earlier}}<earlier_meeting id="{{attr .ID}}">
Title: {{attr .Title}}
{{with .When}}When: {{attr .}}
{{end}}{{with .Attendees}}Attendees: {{range $i, $a := .}}{{if $i}}, {{end}}{{attr $a}}{{end}}
{{end}}{{with .Points}}Points:
{{range .}}- {{attr .Text}}{{with .Who}} ({{attr .}}){{end}}
{{end}}{{end}}{{with .Decisions}}Decisions:
{{range .}}- {{attr .}}
{{end}}{{end}}{{with .Actions}}Action items:
{{range .}}- {{with .Who}}{{attr .}}: {{end}}{{attr .What}}{{with .By}} (by {{attr .}}){{end}}{{with .DueDate}} (due {{.}}){{end}}{{if eq .Todo "suggestion"}} [suggested, not made a to-do]{{else if .Todo}} [to-do {{.Todo}}]{{end}}
{{end}}{{end}}{{with .Open}}Open questions:
{{range .}}- {{attr .}}
{{end}}{{end}}{{with .Projects}}Projects:
{{range .}}- <project slug="{{attr .Slug}}">{{attr .Name}}</project> ({{.Status}})
{{end}}{{end}}</earlier_meeting>
{{else}}(none with minutes)
{{end}}
{{end}}{{if .Question}}The question, typed on the phone:
<question>{{attr .Question}}</question>{{else}}No question was typed: the user tapped at [{{.HeardAt}}]. The lines said in the 45 seconds before the tap:
<heard>
{{range .Heard}}[{{.At}}] {{attr .Text}}
{{end}}</heard>{{end}}
© nwverse · MIT · nwverse/persona