About portable-memory-core
A skill for LLM workflows that centralizes preferences and session state in a portable `.agent/` directory. It loads shared memory and identity rules across different harnesses, and records session history and relevant workspace variables.
Key features
- 01Load preferences and memory from `.agent/`
- 02Pass state across context switches
- 03Apply shared identity rules
- 04Record session summaries
- 05Export relevant workspace variables
Use cases
- 01Carry preferences between LLM runners
- 02Resume state after switching tools
- 03Apply consistent writing and code rules
Install
Copy the command and run it in a terminal.
$
npx skills add https:/After installing, type/portable-memory-corein Claude Code to confirm it loaded.
Bundled filesREADME.md
Details
From the project READMEPortable Memory Core
"Never explain yourself twice."
This skill enforces the One Brain protocol. Instead of scattering your preferences across dozens of system prompts, it centralizes them into a portable .agent/ directory that any LLM runner can hook into.
Core Directives
- Memory Synchronization: Upon startup, read from
.agent/memory.yaml. - Context Passing: Whenever the user switches contexts (e.g., from Cursor to OpenClaw), this skill ensures the new context loads the latest state from
.agent/state.json. - Identity Enforcement: Enforces the
IDENTITY.mdrules (e.g., "Caveman mode", "No em dashes", "Substack voice") universally across all generated code and text. - Harness Agnostic: Works seamlessly whether the upstream harness is Claude CLI, a custom Python script, or a GUI.
Usage
When activated, you will automatically:
- Load
.agent/preferences.yaml - Update
.agent/history.jsonwith the current session summary. - Export relevant variables to the local workspace.
Comments
Markdown supported