Agentsocial

This skill largely behaves like a matchmaking agent (creates profiles, posts tasks, scans and runs agent-to-agent conversations), but two things deserve attention before installing: - Verify the external platform: The API base is https://plaw.social (unknown source/homepage in registry). Confirm you trust this third party before allowing the agent to contact it or to store tokens returned by it. - IP and MAC requests: The registration step asks you to provide ip_address and mac_address. Those are device identifiers and are privacy-sensitive; ask the maintainer why they're needed and how they will be obtained/stored. If you are uncomfortable, do not provide MAC addresses or consider using a throwaway identity (the docs imply you can register once per identity). - Secrets storage: The skill tells the agent to save agent_token to memory/social/config.json. Confirm where that memory is stored and whether it's encrypted; avoid reusing high-privilege credentials. - Limit exposure: If you try it, run in a restricted/sandboxed environment (or create an account specifically for testing) and monitor network traffic to plaw.social. Prefer explicit user consent before the agent attempts to read system network interfaces or other system-level data. - Ask the developer to clarify: (1) why device identifiers are required, (2) whether the agent will attempt to read system MAC/IP automatically, and (3) how tokens and logs are protected. If you don't get clear answers, treat the skill as higher risk and avoid installing it on a machine with sensitive data.