Lancedb Memory
v1.0.0Manage and retrieve long-term memories with LanceDB using semantic vector search, category filtering, and detailed metadata storage.
⭐ 7· 4.9k·37 current·39 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
Files implement LanceDB-based memory storage and search (consistent with the skill name). However the code repeatedly hardcodes an absolute developer path (/Users/prerak/clawd/...), appends that path to sys.path, and uses multiple, inconsistent table names across files (memory, memories, clawdbot_memory). Those hardcoded paths and path-manipulation are not necessary to provide LanceDB memory functionality and are unexpected.
Instruction Scope
The provided code (SKILL.md and Python modules) will create directories and files on disk at a hardcoded path and will modify sys.path to include the developer's local directory. Modifying sys.path to point at an absolute developer path grants the skill the ability to import arbitrary local modules if that path exists on the host — this expands scope beyond just reading/writing a local DB and is a risk.
Install Mechanism
There is no install spec (instruction-only), but the code imports third-party packages (lancedb, pyarrow, pandas) that are not declared anywhere. Lack of dependency declarations or installation instructions is a usability/maintenance problem and can lead to ad-hoc installs later; however there is no automatic download/install step in the package itself.
Credentials
The skill does not request environment variables or credentials (good). Still, it persists data to disk (hardcoded path) which can leak sensitive stored memory if users are unaware; the code does not encrypt or otherwise protect persisted data.
Persistence & Privilege
The skill will create and write a LanceDB store on disk at a developer-specific path and persists memory entries. While always:false (not force-installed), the combination of persistent disk writes and sys.path modification increases blast radius if run in a shared environment. The skill does not modify other skills or system configs, but the hardcoded paths and writing behavior are noteworthy.
What to consider before installing
Key things to consider before installing: 1) Missing description and no declared dependencies — the package imports lancedb, pyarrow, pandas but provides no install instructions; expect to manually install these or the skill may fail. 2) Hardcoded absolute paths (/Users/prerak/...) and sys.path.append calls are unexpected: they will create files under the developer's path and attempt to import modules from that location if it exists on your system. That can lead to accidental imports of local code or surprise file creation. 3) The code persists user 'memories' to disk in plaintext at a fixed path — sensitive data could be stored locally without encryption. 4) The repository contains multiple slightly different implementations (different table names and schemas), suggesting this is a collection of drafts rather than a polished skill. Recommended actions: a) Ask the author to provide a brief description, a dependency list (and/or install spec), and to make the DB path configurable (preferably using an environment variable or XDG_DATA_HOME) rather than hardcoding /Users/…; b) Request removal of sys.path manipulation and instead use relative imports or a proper package layout; c) Audit the code yourself in a sandboxed environment before enabling it for an agent with access to your files; d) If you proceed, run the skill in an isolated container or VM and ensure any persisted memory is stored in a location you control and encrypted if it may contain sensitive data. Providing the above fixes or clarifications would raise confidence and could change the assessment to benign.Like a lobster shell, security has layers — review code before you run it.
latestvk97b77v41q55sypdf1k1e61ymh7zzn8p
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
