Hle Benchmark Evolver

• Analyzed: May 12, 2026, 4:07 AM
• Findings: 1
• Version: 1.0.0
• Hash: 71e687ae3f2d47b9c25606c7d646cafaa11a030973559c161959e26bc722bcac
• Source: palm
• External report: View on VirusTotal

Findings (1)

The skill is highly suspicious due to a critical Remote Code Execution (RCE) vulnerability. The `SKILL.md` documentation explicitly instructs the OpenClaw agent to execute arbitrary shell commands via the `--eval_cmd` parameter. The `run_pipeline.js` script then directly implements this by using `child_process.spawnSync('bash', ['-c', command])` to execute the provided command, which includes a `{{report}}` placeholder that could also be leveraged for further injection if the report path is attacker-controlled. While this capability is presented as a feature to integrate external evaluators, it represents a severe prompt injection and shell injection risk, allowing an attacker to execute arbitrary code on the host system if they can control the input to this skill.