Morgana Mordred Security Sandbox

Educational security training sandbox for AI agents. Contains 5 intentionally vulnerable systems with annotated vulnerability descriptions and tested patches...

This skill appears to implement an offline security lab, but it contains several implementation choices that require caution: - Do NOT run these scripts on a production host or in a privileged environment. The test runner and vulnerable systems execute shell commands, eval(), and arbitrary file reads/writes. Run only inside an isolated VM or Docker container with no access to sensitive mounts. - Inspect and change SANDBOX_PATH in src/mordred_runner.py before running. It is hardcoded to /media/ezekiel/Morgana/sandbox/mordred and will create logs/results there; change it to a path you control (or make it relative/configurable). - Review and remove any embedded secrets or tokens (data_leak and flawed_auth include test 'SECRETS' and SESSION_TOKEN). Treat them as test data, but consider whether they might be mistaken for real credentials. - The 'vaccines' include code that manipulates builtins.__dict__ (vaccine_weak_sandbox). Running that code affects the running Python process. Ensure tests restore state (they attempt to) and run them in isolation. - If your goal is to evaluate agent behavior, prefer spinning up the sandbox inside a disposable container image that has no network mounts and limited privileges. Verify no network calls are made and inspect subprocess invocation carefully. - If you lack the ability to run in an isolated environment, do not install or execute this skill. If you proceed, consider auditing every file (especially modules that call subprocess, open files, or use eval/exec) and patching mordred_runner to use relative paths and to require an explicit --outdir or --sandbox-dir parameter.

Type: OpenClaw Skill Name: morgana-mordred-security-sandbox Version: 2.0.0 The bundle is a security training sandbox ("Mordred") containing intentionally vulnerable Python scripts (SQL injection, RCE, race conditions, and data leaks) along with corresponding patches called "vaccines." While the stated purpose is educational and defensive, the bundle includes functional high-risk vulnerabilities, such as arbitrary code execution via eval() in 'src/systems/weak_sandbox.py' and shell command execution in 'src/mordred_runner.py'. Per the provided guidelines, the presence of these intentional vulnerabilities and risky capabilities, despite the lack of clear malicious intent or data exfiltration, requires a suspicious classification.