The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's Python code in references/auth-bypass-and-xmlrpc.md and references/groovy-rce-exploits.md uses subprocess.run(shell=True) to execute system commands. These commands are often constructed from user-provided input, posing a significant risk of local command injection.
[EXTERNAL_DOWNLOADS]: Instructions in references/auth-bypass-and-xmlrpc.md direct users to download ysoserial-all.jar from a non-trusted GitHub repository (github.com/frohoff) and execute it locally via Java, bypassing safe dependency management.
[REMOTE_CODE_EXECUTION]: The skill provides scripts to perform remote code execution on targets, including reverse shell payloads and Groovy expression injection. While intended for exploitation of targets, the underlying local execution logic is highly dangerous for the environment running the agent.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface. It fetches content from external URLs and extracts data using regex (e.g., extract_result in references/groovy-rce-exploits.md) without validation. 1. Ingestion points: resp.text from requests calls in multiple scripts. 2. Boundary markers: Absent. 3. Capability inventory: subprocess.run, requests.post. 4. Sanitization: The skill does not validate or sanitize extracted data before it is presented to the agent or used in further operations.

ofbiz-exploit