Aleister Agent System

• Analyzed: May 11, 2026, 5:18 AM
• Findings: 1
• Version: 1.0.1
• Hash: 47c513c1d685209a0890887f2316bc87fdcb4aa548d70e3d9eb08b0816b520e4
• Source: palm
• External report: View on VirusTotal

Findings (1)

The analysis of the OpenClaw AgentSkills skill bundle reveals a well-structured multi-agent system with extensive documentation and operational scripts. The primary concern identified is a potential shell injection vulnerability in one of the utility scripts. The `ops/generate-prd.sh` script takes user-supplied arguments (`ISSUE_ID`, `ISSUE_TITLE`) and directly embeds them into `sed` commands for in-place string replacement within a file. If the `ISSUE_TITLE` variable, for example, were to contain shell metacharacters and was not properly quoted by the calling context (e.g., another script or the AI agent itself), it could lead to arbitrary command execution. While the provided example usage `"$ISSUE_TITLE"` suggests quoting, the script itself does not sanitize or escape the input, making it vulnerable to shell injection if misused. This is a vulnerability, not proof of intentional malice. All other scripts and documentation appear to be aligned with the stated purpose of managing an AI agent system, including memory consolidation, cost tracking, system self-checks, and external integrations (e.g., Telegram, Twitter, ElevenLabs). The `AGENTS.md` and `SOUL.md` files contain explicit instructions for the AI agent to prioritize security, avoid data exfiltration, and ask for permission before external actions, which are positive security indicators. The use of `tmux` for persistent sessions and `curl` for external API calls are legitimate operational patterns. The `ops/session-lifecycle-cron.sh` file, as provided, contains a syntax error