The Agent Skills Directory

[SAFE]: The skill provides a legitimate environment for game design using the 'Thermite' methodology. It defines design pillars, personas, and structured workflows for the agent to follow.
[COMMAND_EXECUTION]: The skill includes several Python scripts (add_decision.py, check_mvd.py, export_artifacts.py, new_session.py) intended for execution by the agent. These scripts perform routine file operations such as reading, writing, and zipping local markdown files within the project directory. Analysis of the source code confirms they use standard Python libraries and do not perform any hidden or dangerous commands.
[EXTERNAL_DOWNLOADS]: No external dependencies, remote scripts, or network-based downloads are initiated by the skill. All operations are confined to the local filesystem.
[DATA_EXFILTRATION]: There are no patterns suggesting data exfiltration. The scripts do not use network libraries like requests or urllib, and there are no suspicious curl or wget commands identified in the markdown instructions.
[PROMPT_INJECTION]: The 'Hookify' rules in references/hookify_rules.md and the 'Red Flags' in SKILL.md are safety-oriented guidelines for the design process itself (e.g., preventing scope creep or poor design choices). They do not contain instructions aimed at bypassing the LLM's core safety filters or overriding system instructions.

thermite-design