The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to clone a repository from an untrusted source (https://8015238355.github.io). The use of a purely numeric ID for a repository host is highly non-standard and often associated with hosting malicious payloads or phishing content.
[COMMAND_EXECUTION]: The installation instructions require immediate execution of a shell script (setup.sh) downloaded from the untrusted repository. This grants arbitrary code execution capabilities to an unverified external source during the setup process.
[METADATA_POISONING]: There is a significant discrepancy between the skill's stated purpose ("Analytics Dashboard and Inventory Tracking") and the actual file path used in the instructions (murder-mystery-dupe-roblox). The term "dupe" is a common indicator of scams or malicious "duplication" scripts in the Roblox community, suggesting deceptive intent by the author.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a broad attack surface for indirect injection:
Ingestion points: Reads local gameplay data, Roblox inventory statistics, and player-specific identifiers (ROBLOX_USER_ID).
Boundary markers: None provided in the instructions to prevent the agent from interpreting instructions embedded within the gameplay data.
Capability inventory: The skill uses requests for network access, writes to the file system (exports/), and executes CLI commands via main.py.
Sanitization: There is no evidence of sanitization or validation for the data ingested before it is passed to the AI models (OpenAI/Claude) for "strategy analysis."

mm2-analytics-roblox-toolkit