Skip to main content
prompt GP ChatGPT

task_memory_step — Claude Code prompt: what it does, how to install

The task_memory_step Claude Code prompt retrieves a record of one step from an earlier task. It shows the step’s actions and the text it read at the time.

Apache-2.0 ★ 9,488 updated 2026-11-16

Install into Claude Code · Manually: put the files in place All methods and platforms ↓
.claude/commands/task_memory_step.md
Open on GitHub

What task_memory_step does

The prompt looks up a past task step using the task’s exact ID and the step number from task_memory_timeline. Its record contains the actions taken and text read from a screen, page, or command output.

Sensitive data has already been redacted, and the captured text is shortened. The record reflects an earlier moment; read the live screen again to see what is true now.

What it helps with

  • retrieve the actions of a specific past task step
  • recover text a step read earlier
  • separate a past screen state from the current one

Example prompts for the agent

  • Show me step 2 from task abc123.
  • What text did step 4 of task abc123 read?
  • Retrieve the actions from step 1 of task abc123.

How to install task_memory_step

Manually: put the files in place

.claude/commands/task_memory_step.md

Questions about task_memory_step

How do I install task_memory_step in Claude Code?

Add the prompt file at .claude/commands/task_memory_step.md.

What does task_memory_step do?

It retrieves a past task step’s record, including its actions and the text it read at that time.

Does task_memory_step show the current screen state?

No. The record captures an earlier state. Read the live screen again to see what is happening now.

Source text

task_memory_step.md · Apache-2.0
Read the raw record of one step of a past task (an exact id plus the step number from task_memory_timeline): what the step did and the text it read then (screen, page, command output, already redacted and truncated). It shows what was true at that time; re-read the live screen when the current state matters.

© wangzan101 · Apache-2.0 · wangzan101/Mika

Similar prompts

All prompts for Agent setup: memory, context, orchestration →

From the same set132

The whole set →
phone_call_state The phone_call_state Claude Code prompt reads cellular call state, local AI calling readiness, and available SIM IDs. It distinguishes dialing and connecting from a completed call. GP ChatGPT ★ 9,488repo stars phone_call_hangup The phone_call_hangup Claude Code prompt ends the active cellular call when the user asks to hang up. It checks the call state and uses the current call_id. GP ChatGPT ★ 9,488repo stars user_screen The user_screen Claude Code prompt answers questions about a phone screenshot. It sticks to details that are clearly visible and gives no advice or steps. GP ChatGPT ★ 9,488repo stars task_memory_timeline The task_memory_timeline Claude Code prompt displays the timeline of one past task by its exact ID. It lists the goal, conclusion, apps and sites, plus brief step labels with numbers and times. GP ChatGPT ★ 9,488repo stars mika_login_site The mika_login_site Claude Code prompt opens a site in Mika’s browser so the user can sign in themselves. The agent waits for approval before opening the page, then calls mika_confirm_login after the user finishes. GP ChatGPT ★ 9,488repo stars skill_read skill_read — a prompt for Claude Code and ChatGPT that retrieves a skill’s full text by name. It can return the main text or an attachment listed in the skill’s files. GP ChatGPT ★ 9,488repo stars screenshot The screenshot Claude Code prompt captures the current screen as an image. Save it to a new PNG path under /workspace, then call share_file to deliver it. GP ChatGPT ★ 9,488repo stars knowledge_read The knowledge_read prompt for Claude Code and ChatGPT reads information about apps and websites. It opens an attachment or looks up a resource that has not been opened yet. GP ChatGPT ★ 9,488repo stars

Project sponsors

Companies that keep this analytics open