The Agent Skills Directory

[NO_CODE]: The skill primarily consists of instructional Markdown files and configuration data. It provides guidelines for decision-making, such as classifying constraints and verifying results.
[COMMAND_EXECUTION]: The skill includes Python scripts within its documentation as worked examples. These scripts (e.g., snake_rename.py and sample_size.py) are benign. The renaming script implements a safety-first approach with a mandatory flag for destructive actions and uses standard libraries (os, re, sys). The sample size calculator is a pure mathematical utility using the math library.
[DATA_EXFILTRATION]: No network operations, credential harvesting, or unauthorized data access patterns were found. The skill explicitly instructs agents not to invent facts or fill unknown contracts with defaults, which reduces the risk of generating misleading or potentially harmful artifacts.
[PROMPT_INJECTION]: The instructions focus on reinforcing the agent's adherence to user priorities and rigorous proof standards. There are no attempts to bypass safety filters or override system-level constraints.
[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md and referenced files do not contain any dynamic command injection patterns (such as ! followed by backticks). All shell examples provided are static documentation of expected behavior.

starting-point