LeoYeAI/agent-master-skills / skills/agent-memory-system-guide
Agent Memory System Guide
VerifiedAn agent memory workflow guide for OpenClaw and Codex with MEMORY.md, daily notes, SESSION-STATE, working-buffer, Obsidian archiving, and optional OpenViking support.
docs/superpowers/specs/2026-03-27-memory-positioning-design.md
View on GitHub# Memory Positioning and Boundary Design
## Scope
This design tightens the repository messaging around what this project is and is not.
The update is intentionally documentation-only:
- Clarify that this repository is a local-first memory workflow, not a hosted memory platform.
- Make the file-layer boundaries easier to understand at a glance.
- Introduce lightweight concepts for stable profile, active context, and optional project scope.
- Reframe OpenViking and similar systems as optional recall backends or adapters.
This iteration does not add new commands, new storage formats, or automatic extraction behavior.
## Goals
- Preserve the repository's current local-file workflow as the core product.
- Explain the difference between recovery state, distilled memory, raw notes, and archive material.
- Make room for future semantic recall integrations without turning them into hard dependencies.
- Incorporate lessons from memory-platform products without changing this repository into one.
## Non-Goals
- Build a hosted memory API.
- Add connector infrastructure, sync daemons, or MCP services.
- Introduce automatic profile extraction, conflict resolution, or forgetting logic.
- Replace the current templates or script contract.
## Current Problem
The current README documents the workflow correctly, but the product positioning is still easy to misread:
- `MEMORY.md`, `memory/`, Obsidian, and optional semantic search are described, but their boundaries are spread across multiple sections.
- OpenViking is marked optional, but the broader concept of "optional recall backend" is not explicit.
- The repository does not clearly say that it is a workflow guide and file contract rather than an online memory product.
## Options Considered
### Option 1: Keep the current README structure and only mention more optional tools
Pros:
- Smallest edit.
Cons:
- Does not clarify the product boundary.
- Leaves memory-layer concepts implicit.
### Option 2: Add explicit memory-layer and backend-boundary sections
Pros:
- Matches the repository's actual value.
- Makes external references such as Supermemory easier to position correctly.
- Improves onboarding without changing runtime behavior.
Cons:
- Requires moderate README restructuring across languages.
### Option 3: Expand into a larger architecture document and defer README cleanup
Pros:
- More detailed.
Cons:
- Adds documentation weight before fixing the main entry points.
## Recommended Design
Adopt Option 2.
### README messaging changes
Add explicit language that this repository is:
- A local-first Agent memory workflow
- Based on durable files and manual distillation
- Compatible with optional semantic recall backends
- Not a hosted or API-first memory platform
### Memory boundary section
Add a compact section that explains:
- `SESSION-STATE.md`: active recovery state for the interrupted task
- `working-buffer.md`: rough scratchpad and write-ahead notes
- `MEMORY.md`: distilled stable profile, preferences, decisions, and recurring pitfalls
- `memory/`: daily notes and project-scoped raw history
- Obsidian or external search backends: deeper archive and optional recall
### Stable profile and active context
Add guidance that `MEMORY.md` should emphasize:
- Stable profile: lasting preferences, conventions, recurring decisions
- Active context: only the small amount of ongoing context that should survive startup
Daily notes remain the place for volatile execution history.
### Project scope concept
Add a lightweight recommendation that entries may include a project or repo reference when one workspace serves multiple efforts.
This stays a convention, not a new required schema.
### Optional backend positioning
Broaden OpenViking messaging so it still remains optional, while making clear that:
- OpenViking is one example of an optional semantic recall backend
- Other tools or services can fill the same role later
- The local recovery layer remains authoritative
## Files Expected To Change
- `README.md`
- `README_EN.md`
- `README_CN.md`
- `INSTALL.md`
## Acceptance Criteria
- The main README entry points clearly describe the repository as a local-first workflow.
- The documentation distinguishes recovery state, distilled memory, raw notes, and archive/search layers.
- OpenViking remains explicitly optional and not required for the core workflow.
- The documentation allows optional external backends without implying a platform pivot.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |