Agent Template

Bot-vs-bot parimutuel prediction markets on Solana. Trade real creator attention metrics.

Key points to consider before installing or running: - Do not run the recommended 'curl https://app.twzrd.xyz/raw/wzrd-trade.sh | bash' unless you (or someone you trust) have inspected that script line-by-line. Piping remote scripts to a shell executes arbitrary code on your machine. - The code requires an Ed25519 private key (WZRD_PRIVATE_KEY). Never supply your primary wallet/private key. Use an ephemeral or dedicated wallet with only the minimal SOL required (the service requires ~0.001 SOL) and minimal funds so a compromised key has limited impact. - The registry metadata lists no required env vars, but the example_agent and README clearly require WZRD_PRIVATE_KEY. Ask the publisher to correct metadata before trusting automation. - Audit the included example_agent.py and any remote scripts for hidden exfiltration (HTTP calls to unexpected hosts, uploading files, or reading unrelated local files). The provided example_agent.py does standard API calls and local env parsing, but you should still inspect any downloaded script. - Prefer cloning the GitHub repo and reviewing code locally (git clone https://github.com/twzrd/twzrd-agent-template) rather than executing a remote installer. Run the agent in an isolated environment (container or VM) and network monitor the outbound calls to verify only expected endpoints (api.twzrd.xyz and app.twzrd.xyz) are contacted. - If you need autonomous operation, limit the agent's privileges and funds, and consider manual approval for sensitive actions (redeem/relay/transaction-signing). Ask the publisher to provide a signed release or reproducible build and to fix registry metadata so required secrets are declared explicitly.

Type: OpenClaw Skill Name: wzrd Version: 0.5.0 The skill bundle is classified as suspicious primarily due to a significant prompt injection vulnerability identified in `README.md`. The quickstart instructions include `curl -fsSL https://app.twzrd.xyz/raw/wzrd-trade.sh | bash`, which allows an AI agent to execute an arbitrary remote script if instructed to follow installation guides. While the `example_agent.py` code itself appears to be a legitimate prediction market agent, using the provided private key only for authentication to `api.twzrd.xyz`, the presence of a `curl | bash` instruction in agent-facing documentation represents a high-risk capability that could lead to remote code execution if the `wzrd-trade.sh` script or its hosting server were compromised, or if the agent is maliciously prompted to execute it.