Engram Memory

What to check before installing or running this skill: - Do not run setup scripts or docker-compose in a sensitive production environment until you audit them. The provided setup pulls Docker images (engrammemory/fastembed) from an external registry; pull/rebuild images locally from the included Dockerfile if you want to avoid running untrusted images. - Review scripts/setup.sh and docker-compose.yml to see exactly what containers and network ports will be created. Prefer running in an isolated VM or disposable machine the first time. - Inspect docker/fastembed/Dockerfile and the fastembed service code for any unexpected network calls or telemetry. The README claims 'no phone‑home', but confirm by grepping the repo for outbound HTTP requests and external domains (e.g., engrammemory.ai) before trusting it. - If you are concerned about privacy, disable autoCapture and autoRecall in the plugin config (set autoCapture=false and autoRecall=false) or limit them until you’re comfortable with behavior; check where conversation content is stored (Qdrant collection) and how retention/forgetting works. - Audit mcp/server.py and any server scripts for open network bindings; if you expose an MCP server, bind to localhost or restrict access via firewall. - If you lack the expertise to audit containers and Python/JS code, run the stack in an isolated environment (VM) and observe network egress (e.g., with a network monitor) to ensure nothing phones home. Taken together: the functionality matches the stated purpose, but because the skill will pull and run third‑party containers and can automatically persist conversation content into a local store, treat it as potentially risky until you've inspected or rebuilt the images and confirmed no unexpected external communications.