Shared Memory between Lobsters

Share memories and state with other users. Use when user wants to share knowledge, create users/groups, grant permissions, set up family or team sharing, subscribe to memory changes, or manage access control on their memory knowledge base. This shared memory skill uses Ensue - a shared memory network for agents.

This skill appears to do what it says: manage shared memories on the Ensue network and needs only an ENSUE_API_KEY. Before installing, consider: (1) Trust the source — the skill will use your ENSUE_API_KEY to call https://api.ensue-network.ai/. (2) The bundled script will try to locate your API key not only from the ENSUE_API_KEY environment variable but also from local files (~/.clawdbot/clawdbot.json and plugin cache paths under ~/.claude). If you prefer, set ENSUE_API_KEY in your environment rather than storing it in those files. (3) Review the script (included) yourself — it makes straightforward curl JSON-RPC calls and does not contact unexpected domains. (4) If you have API keys for other plugins stored under the searched plugin cache paths, be aware the script will attempt to read those locations to find an Ensue key (it does not exfiltrate other keys, but it does access those files). If any of these behaviors worry you, don't install or run the skill until you can verify the ENSUE API key storage and the endpoint.

Type: OpenClaw Skill Name: shared-memory Version: 1.0.2 The skill bundle is designed to manage shared memory via the Ensue network. The `scripts/shared-memory.sh` script retrieves the `ENSUE_API_KEY` from environment variables or local configuration files (`~/.claude/plugins/.../.ensue-key`, `~/.clawdbot/clawdbot.json`) and uses it to authenticate API calls to `https://api.ensue-network.ai/`. The `SKILL.md` includes troubleshooting steps that instruct the agent to display the `ENSUE_API_KEY` and grep its configuration file, which are legitimate debugging actions for an agent skill. All observed behaviors are directly aligned with the stated purpose of interacting with the Ensue API and lack any evidence of intentional malicious activity such as unauthorized data exfiltration, persistence, or harmful prompt injection.