← Back to Skills Marketplace
Agent Memory Kit
by
ryancampbell
· GitHub ↗
· v2.1.0
1848
Downloads
1
Stars
5
Active Installs
1
Versions
Install in OpenClaw
/install agent-memory-kit
Description
Provides a structured AI agent memory system separating episodic, semantic, and procedural memories to preserve knowledge and processes over time.
README (SKILL.md)
Agent Memory Kit
Type: Practice / Framework Dependencies: None (markdown only)
Description
A structured memory system for AI agents. Prevents the "forgot how to do things" problem by separating memory into episodic (what happened), semantic (what I know), and procedural (how to do things) layers.
Installation
# Create memory folder structure
mkdir -p memory/procedures
# Copy templates
cp templates/ARCHITECTURE.md memory/
cp templates/feedback.md memory/
cp templates/procedure-template.md memory/procedures/
Usage
- Read
README.mdto understand the system - Add memory loading to your wake routine (AGENTS.md)
- Use templates when logging events, creating procedures, tracking feedback
Files
| File | Purpose |
|---|---|
README.md |
Full documentation |
templates/ARCHITECTURE.md |
Memory system overview (copy to memory/) |
templates/feedback.md |
Success/failure tracking template |
templates/procedure-template.md |
How-to document template |
templates/daily-template.md |
Daily log template |
templates/compaction-survival.md |
NEW: Pre-compaction flush guide |
templates/context-snapshot-template.md |
NEW: Quick context save template |
helpers/check-compaction.sh |
NEW: Token limit checker |
Key Concepts
- Episodic memory: Daily logs of WHAT happened
- Semantic memory: Curated knowledge (MEMORY.md)
- Procedural memory: HOW to do things
- Feedback loops: Learn from success/failure
The Rule
Always capture the HOW, not just the WHAT. Future-you needs the steps.
Usage Guidance
This kit is largely local markdown + shell helpers and appears coherent with its stated purpose. Before installing or running anything: 1) Inspect the two shipped shell scripts (lib/search.sh and helpers/check-compaction.sh) to ensure they do not call unexpected network endpoints or run arbitrary commands; 2) If you follow the README's git instructions, confirm the remote origin is trusted (the skill metadata says 'source: unknown' even though README points to a GitHub repo); 3) Adding the bin path to your shell rc will run this tool more conveniently but modify your shell startup — make that change manually so you can review it; 4) Avoid blindly running `git pull` inside the skill folder unless you trust the remote; consider cloning directly from an official repo URL you verify. If you want extra safety, run the CLI in a sandboxed environment or inspect the code fully before adding it to your PATH or your agent wake routines.
Capability Analysis
Type: OpenClaw Skill
Name: agent-memory-kit
Version: 2.1.0
The OpenClaw AgentSkills bundle provides a memory management and search system for AI agents. It uses shell scripts (`lib/search.sh`, `helpers/check-compaction.sh`) to operate on local markdown files within the agent's workspace, employing standard Unix utilities like `grep`, `find`, and `jq` for its stated purpose. While the markdown files contain instructions for the agent (prompt injection) to modify its environment (e.g., `PATH`), integrate with its `AGENTS.md` and `HEARTBEAT.md`, and manage its memory files, these actions are transparently aligned with the benign goal of improving agent self-management and recall. There is no evidence of data exfiltration, malicious execution, persistence, or obfuscation beyond the skill's legitimate functionality.
Capability Assessment
Purpose & Capability
The skill's name and SKILL.md describe a local memory system and search/recall CLI. The repo includes markdown templates and shell scripts (search implementation and a compaction checker), which are expected for a CLI-based search/tooling kit. Required env vars and binaries are none, which matches a mostly-markdown + shell implementation.
Instruction Scope
Runtime instructions are limited to copying templates into a local memory/ folder, adding the bin path to the shell PATH, and optionally running a local token-check script. That's within scope. One instruction (INSTALL-SEARCH.md) suggests running `git pull origin main` inside the skill directory — this can fetch remote code and is out-of-band relative to purely local template install; treat it as a remote update step to verify before executing.
Install Mechanism
There is no automated install spec — the skill is instruction-only. That lowers risk. The README suggests cloning from GitHub (a normal install path). No downloads from untrusted URLs or arbitrary extract operations are present in the provided docs.
Credentials
The skill declares no required environment variables, credentials, or config paths. The instructions ask only to copy templates and optionally add a local bin path to the user's shell config — permissions requested are proportional to the stated purpose.
Persistence & Privilege
The skill is not marked always:true and does not request elevated agent privileges. It does recommend adding its bin directory to your shell PATH and adding a wake-routine step (AGENTS.md) — these are user-controlled changes that increase the tool's convenience but require you to edit shell or agent routines. Be aware these edits change your shell startup behavior (not a platform-level privilege escalation).
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install agent-memory-kit - After installation, invoke the skill by name or use
/agent-memory-kit - Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.1.0
- Added new templates: compaction survival guide and context snapshot template.
- Introduced a token limit checker script.
- Updated documentation to reflect new memory management tools.
- Improved instructions for installation and usage of memory templates.
- Clarified separation of episodic, semantic, and procedural memory layers.
Metadata
Frequently Asked Questions
What is Agent Memory Kit?
Provides a structured AI agent memory system separating episodic, semantic, and procedural memories to preserve knowledge and processes over time. It is an AI Agent Skill for Claude Code / OpenClaw, with 1848 downloads so far.
How do I install Agent Memory Kit?
Run "/install agent-memory-kit" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Agent Memory Kit free?
Yes, Agent Memory Kit is completely free (open-source). You can download, install and use it at no cost.
Which platforms does Agent Memory Kit support?
Agent Memory Kit is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Agent Memory Kit?
It is built and maintained by ryancampbell (@ryancampbell); the current version is v2.1.0.
More Skills