Abaddon

• Analyzed: May 12, 2026, 4:24 AM
• Findings: 1
• Version: 1.0.0
• Hash: 0ff77aee8e3087b1152c6b02b750476b6f861fe87eb7a57cdccaaecdd716f356
• Source: palm
• External report: View on VirusTotal

Findings (1)

```json { "classification": "suspicious", "summary": "The skill is designed for a 'red team security audit,' instructing the OpenClaw agent to 'think like an adversary' and perform extensive system introspection. This includes scanning for API keys and secrets in files, environment variables, and Git history (`grep -r 'sk-\|xai-\|Bearer\|api_key'` and `git log -p --all | grep -i 'password\|secret\|api_key\|token'` in `templates/abaddon-prompt.md`), checking sensitive file permissions, and modifying the OpenClaw cron schedule (`setup/cron-seed.sh`). While the stated intent is to find vulnerabilities and report summaries to a security topic, the explicit prompt injection instructing the agent to adopt an 'active attacker' mindset, combined with broad access to sensitive data and system configuration, represents significant risky capabilities that could be easily repurposed or misinterpreted if the agent's guardrails were to fail, making it suspicious rather than benign." } ```