ClawHub

Memorine

v0.1.0

Human-like memory for AI agents. Facts, events, procedures, contradiction detection, forgetting curve, and cross-agent sharing. Pure Python + SQLite.

⭐ 0· 208·1 current·1 all-time
by@osvfelices
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
medium confidence
✓
Purpose & Capability
Name/description, required binaries (python3), and declared MCP tools align with a local Python-based memory service. No unrelated credentials, binaries, or config paths are requested.
ℹ
Instruction Scope
SKILL.md limits activity to running an MCP server and storing data in ~/.memorine/memorine.db. It does not instruct reading other system files or network exfiltration. However the service creates and uses a shared DB file (by default in the user's home) — that is expected for cross-agent sharing but is a privacy/persistence surface you should consider (permissions, backups, who can read the file).
ℹ
Install Mechanism
Install spec uses an 'uv' package installer entry (and README shows pip install memorine). This is not a download-from-arbitrary-URL install; it expects a standard Python package. The slight mismatch (uv vs pip) is minor but worth confirming which install path will be used in your environment.
✓
Credentials
No environment variables or external credentials are requested, which is proportionate to a local-memory tool. The main resource access is the filesystem (the DB file). Ensure the claim that agents can only modify their own namespace is actually enforced by the package if you rely on it for isolation.
✓
Persistence & Privilege
The skill is not always-enabled and does not request elevated system privileges. It will install a binary/module and create a persistent DB in the user's home directory — expected for a memory service but represents persistent storage the agent can use.
Assessment
This skill appears coherent for a local Python+SQLite memory MCP server, but you should: (1) verify the package source (pip/uv vs the GitHub repo) before installing; (2) inspect the package code on the referenced GitHub/PyPI to confirm the 'per-agent namespace' enforcement; (3) be aware it will create ~/.memorine/memorine.db (ensure filesystem permissions and backups meet your privacy needs); (4) treat optional embeddings/ui extras as larger installs (models/deps) and only install them if needed; (5) if you have strict isolation requirements, run the MCP server in a constrained environment (dedicated user, container, or VM) so agent processes cannot access other sensitive files.

Like a lobster shell, security has layers — review code before you run it.

latestvk976h93a9q5j17005de4ys83ks82k1sd

License

MIT-0
Free to use, modify, and redistribute. No attribution required.
Termshttps://spdx.org/licenses/MIT-0.html

Runtime requirements

Binspython3

Install

uv
Bins: memorine
uv tool install memorine

Comments