history_read — Claude Code prompt for conversation history
The history_read Claude Code prompt retrieves details from earlier conversation turns. Use it to resume an interrupted task without collecting the same findings again.
Apache-2.0 ★ 9,488 updated 2026-11-16
.claude/commands/history_read.mdWhat history_read does
The prompt retrieves what happened in earlier turns. Without a turn number, it lists them newest first, showing the user's request, the outcome, and when it happened. That gives the agent context for a paused task.
For a selected turn, it can return the outcome, Mika's progress notes, and text read at the time, such as page contents or command output. Results are ordered oldest first. Each call returns about {{page}} characters; pass nextOffset as the offset to read further. Detailed records are kept only for recent tasks. With recorded=false, only the outcome remains. The history reflects an earlier state, so current information requires a fresh read.
What it helps with
- resume a paused task using earlier conversation turns
- find the outcome and time of a previous task
- retrieve Mika's notes and material read during a turn
- continue reading a long record from
nextOffset
Example prompts for the agent
- Find the earlier turn where we stopped and continue that task.
- Show what I asked for and how the latest turn ended.
- Read the selected turn's material, then continue from
nextOffset.
How to install history_read
Manually: put the files in place
.claude/commands/history_read.mdHow to use
Paste the text at the start of a chat or into project instructions
Questions about history_read
How do I install history_read in Claude Code?
Place the command file at .claude/commands/history_read.md.
What does history_read do?
It lists earlier turns or retrieves the outcome of a selected turn, along with Mika's notes and material read at the time.
What details remain for older tasks?
Detailed records are kept only for recent tasks. When recorded=false, only the outcome is available.
Source text
history_read.md · Apache-2.0
Look up what earlier turns of this conversation found, so you can reuse it instead of gathering it again (for example when the user stopped a task and now asks to continue). Without turn: lists the earlier turns, newest first, with what the user said, how it ended and when. With turn: that turn's full outcome, Mika's progress notes and the text it read then (screens, pages, command output), oldest first, about {{page}} characters per call; pass nextOffset as offset to read on. Only the most recent tasks keep this detail (recorded=false means only the outcome is left). It shows what was true then; read again when the current state matters.
© wangzan101 · Apache-2.0 · wangzan101/Mika