The Agent Skills Directory

[SAFE]: No malicious patterns, prompt injections, or exfiltration vectors were detected during the analysis.
[DATA_EXFILTRATION]: The skill operates entirely through standard input and output. It does not perform network requests or access sensitive local files like SSH keys or environment variables.
[COMMAND_EXECUTION]: There are no subprocess calls, shell executions, or dynamic command assemblies present in the code.
[REMOTE_CODE_EXECUTION]: The skill does not download external scripts or execute remote payloads. It relies on a local relative import for protocol validation, which is standard for structured project environments.
[INDIRECT_PROMPT_INJECTION]: While the skill processes external game data (JudgeTask), it uses strict JSON parsing and schema validation (validate_judge_task). Furthermore, it lacks any dangerous capabilities (like file writing or network access) that could be exploited via malicious input data.

werewolf-player