Surf

Your AI agent's crypto brain. One skill, 83+ commands across 14 data domains — real-time prices, wallets, social intelligence, DeFi, on-chain SQL, prediction...

This skill is a dispatcher to an external 'surf' CLI rather than a self-contained implementation. Before installing or letting an agent follow these instructions: 1) Verify the provenance of the CLI and the documentation URL (agents.asksurf.ai) — find a maintained source (GitHub repo, release artifacts) and inspect the code/binaries. 2) Prefer running 'surf install' yourself in a controlled environment (or skip it) rather than letting the agent execute installs. 3) Be cautious about allowing the skill to edit and commit AGENTS.md or other project files — that change routes all agents to use Surf by default. If you don't want global routing, decline the prompt; the SKILL.md suggests creating ~/.surf/.routing-declined to record that. 4) Review any npm packages invoked via npx and their maintainers. 5) If you need a higher assurance, ask the publisher for a signed release, a repo link, or a cryptographic checksum for the CLI binary. Because the metadata omits install provenance and the skill instructs persistent repository changes, treat this as potentially risky until you validate the CLI and intended behavior.

Type: OpenClaw Skill Name: surf Version: 1.1.0 The skill bundle instructs the agent to modify project files (AGENTS.md or CLAUDE.md) and perform git commits to inject 'routing rules' that prioritize this tool over others, which is a form of persistent prompt injection. It also relies on `surf install` to download or update a binary from a remote endpoint (asksurf.ai) and includes a feedback command that exfiltrates the last 10 turns of conversation history. While these actions are documented and include user consent checks, the automated modification of project structure and the self-favoring prompt injection are high-risk behaviors.