Skip to content

LeoYeAI/agent-master-skills / skills/agent-memory-system-guide

Agent Memory System Guide

Verified

An agent memory workflow guide for OpenClaw and Codex with MEMORY.md, daily notes, SESSION-STATE, working-buffer, Obsidian archiving, and optional OpenViking support.

FileKindSize
INSTALL.mdmarkdown1.1 KB
README.mdmarkdown12.4 KB
README_CN.mdmarkdown11.6 KB
README_EN.mdmarkdown12.3 KB
SKILL.mdskill md18.8 KB
_meta.jsonconfig833 B
docs/superpowers/plans/2026-03-21-agent-memory-system-guide-obsidian-native.mdmarkdown5.3 KB
docs/superpowers/plans/2026-03-25-memory-backup-import-export.mdmarkdown3.9 KB
docs/superpowers/plans/2026-03-25-memory-examples-maintenance.mdmarkdown3.2 KB
docs/superpowers/plans/2026-03-25-report-warning-example.mdmarkdown1.9 KB
docs/superpowers/plans/2026-03-26-import-clean-mode.mdmarkdown8.6 KB
docs/superpowers/plans/2026-03-26-metadata-bootstrap-safety.mdmarkdown5.4 KB
docs/superpowers/plans/2026-03-27-memory-positioning.mdmarkdown2.6 KB
docs/superpowers/plans/2026-03-27-report-json.mdmarkdown5.5 KB
docs/superpowers/specs/2026-03-21-agent-memory-system-guide-obsidian-design.mdmarkdown5.1 KB
docs/superpowers/specs/2026-03-25-memory-backup-import-export-design.mdmarkdown3.1 KB
docs/superpowers/specs/2026-03-25-memory-examples-maintenance-design.mdmarkdown2.8 KB
docs/superpowers/specs/2026-03-26-import-clean-mode-design.mdmarkdown5.6 KB
docs/superpowers/specs/2026-03-26-metadata-bootstrap-safety-design.mdmarkdown5.0 KB
docs/superpowers/specs/2026-03-27-memory-positioning-design.mdmarkdown4.4 KB
docs/superpowers/specs/2026-03-27-report-json-design.mdmarkdown4.0 KB
manifest.tomlconfig944 B
scripts/memory_capture.pyscript17.3 KB
templates/OBSIDIAN-NOTE.mdmarkdown530 B
templates/SESSION-STATE.mdmarkdown137 B
templates/memory-capture.mdmarkdown385 B
templates/working-buffer.mdmarkdown105 B
tests/test_gitignore_publish_artifacts.pyscript514 B
tests/test_install_contract.pyscript384 B
tests/test_memory_capture_contract.pyscript6.9 KB
tests/test_memory_capture_script.pyscript16.8 KB
tests/test_memory_recovery_contract.pyscript1.5 KB
tests/test_obsidian_contract.pyscript503 B
tests/test_obsidian_template_contract.pyscript183 B
tests/test_openclaw_memory_search_contract.pyscript366 B
tests/test_openviking_optional_contract.pyscript371 B
tests/test_readme_openviking_optional.pyscript383 B
tests/test_release_manifest_contract.pyscript1.3 KB
tests/test_skill_positioning_contract.pyscript849 B

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.
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: 471kUpdated Claude CodeCodexVerifiedInstall
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: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
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 availableStars: 322kUpdated Claude CodeCodexVerifiedInstall