wbso

SUSPICIOUS: the stated purpose is coherent, but the skill relies on an unreviewable local `wbso` CLI that reads credentials and rich local context, then makes networked API calls. Same-org documentation reduces maliciousness concerns, yet the black-box executable, multi-path CLI discovery, and transitive bundled-skill trust make the overall security posture high-risk rather than benign.

This CLI performs expected WBSO.ai API and local context operations, but parse_args() has a command-injection risk because untrusted flag names are evaluated with eval. Sourced configuration files can also execute arbitrary shell code, and a malicious WBSO_API_BASE_URL could receive the bearer API key. The context command prints local Git information and Claude/Codex user prompts to stdout, but this code does not directly transmit those appended signals. Severity is moderate; no clear evidence of intentional malware or obfuscation is present.