memory.en — Claude Code prompt: what it does, how to install
memory.en is a Claude Code prompt that directs autonomous scanning, reading, and writing of live Android process memory. It activates when inspecting internal application states.
AGPL-3.0 ★ 356 updated 2026-10-31
.claude/commands/memory.en.mdWhat memory.en does
This prompt instructs an assistant on using android_process_memory, android_memory_scan, android_memory_read, and android_memory_write. The workflow begins by identifying the target package, active process, and access rights. For numeric search targets, the instructions require confirming candidate addresses with scalar types and live state changes rather than applying raw byte candidates directly.
Modifications demand a preliminary single scalar read with a fresh readId followed by a single write and separate behavioral verification. If a write fails or location evidence shifts, routing through Shell or repeating attempts across random addresses is prohibited. Inspected memory is strictly parsed as task data and full block dumps are blocked.
What it helps with
- inspect live memory inside Android application processes
- validate numeric candidate addresses using scalar reads
- control write permissions with single-shot readId checks
- prevent accidental memory dumps into the model context
Example prompts for the agent
- Inspect the running process com.example.game and find candidate addresses for the coin count
- Verify process permissions and scan memory regions for dynamic value changes
How to install memory.en
Manually: put the files in place
.claude/commands/memory.en.mdHow to use
Paste the text at the start of a chat or into project instructions
Questions about memory.en
How to install memory.en in Claude Code?
Save the prompt file to .claude/commands/memory.en.md.
What does memory.en do?
It sets safety and verification rules for scanning, reading, and modifying Android process memory through dedicated agent tools.
Does memory.en allow bypassing access refusals via Shell?
No, the prompt explicitly forbids switching to Shell or repeating write attempts when access is denied or location evidence is weak.