python-executor

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). Most links are documentation or assets on inference.sh (and a benign GitHub API/example.com call), but the curl | sh installer at cli.inference.sh plus downloadable binaries on dist.inference.sh (with only SHA‑256 checksums and an unfamiliar registry domain clawhub.ai) represent a supply‑chain risk and should be treated as suspicious unless you fully trust and independently verify the provider.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.65). The required runtime workflow executes user-supplied Python code (the code field) which can fetch arbitrary public web content (e.g., via requests/httpx/scrapy), and any fetched HTML/text could be printed/captured and thus injected into the agent’s LLM context via capture_output/returned outputs.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The Quick Start directs users to execute a remote installer via "curl -fsSL https://cli.inference.sh | sh" (which in turn downloads binaries from dist.inference.sh), so https://cli.inference.sh (and the referenced dist.inference.sh) are runtime-fetched scripts/binaries that are executed and thus constitute a high-confidence external code-execution dependency.