Solanaprox

Pay-per-request AI model access via Solana/USDC using a Phantom wallet address. Query Claude and GPT models without API keys. Wallet-native authentication wi...

This skill appears to do what it says (send your wallet address to solanaprox.com and pay per call), but take precautions before using it: - Do NOT put private keys, seed phrases, or secret keys into SOLANA_WALLET. The SKILL.md intends SOLANA_WALLET to be a public wallet address, but the variable name is ambiguous and a mistake could expose secrets. - Be aware that all prompts and responses are routed through a third-party proxy (solanaprox.com). Do not send sensitive data or private information unless you trust that service. - The doc suggests running `npx solanaprox-mcp` (optional). Running npx will download and execute code from npm — review the package source (npm page and GitHub) before running. - The SKILL.md metadata references a GitHub repo and an npm package; the registry record lacked a homepage. Verify those links independently (inspect repository and package) before trusting the service. - If you want to try it, use a dedicated Solana wallet with minimal funds (e.g., <$5) rather than your main wallet. If you want a firmer benign/malicious determination, provide the package source (GitHub repo or the npm package contents) or clarify exactly what value users are expected to set in SOLANA_WALLET (address vs. key).

Type: OpenClaw Skill Name: solanaprox Version: 2.0.0 The skill is classified as suspicious due to several high-risk behaviors and a transparency issue. Firstly, the `SKILL.md` instructs the agent to execute `curl` commands that directly embed the `$SOLANA_WALLET` environment variable into shell commands without apparent sanitization, creating a shell injection vulnerability (RCE risk) if the agent's execution environment doesn't properly escape inputs. Secondly, the skill instructs the agent to run `npx solanaprox-mcp`, which allows arbitrary code execution by downloading and running an npm package, posing a significant supply chain risk. Lastly, the 'Security Manifest' in `SKILL.md` falsely claims that `https://solanaprox.com/` is the 'only' external endpoint called, while later sections explicitly provide a `curl` command to `https://aiprox.dev`, indicating a lack of transparency.