The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/update-skill.py is designed to download a ZIP archive from the official repository on GitHub and replace the skill's own instructions and executable scripts on the local filesystem.
[PROMPT_INJECTION]: The SKILL.md file instructs the agent to run an automatic update check using update-skill.py at the start of every session and specifically tells the agent to "reopen and reread" the skill if an update is detected, allowing the skill to dynamically modify the agent's operating instructions.
[PERSISTENCE]: The automatic update mechanism, combined with the instruction for the agent to run it on every invocation, establishes a persistent update channel that survives across different agent sessions.
[COMMAND_EXECUTION]: Several utility scripts, including benchmark.py and generate-release-manifest.py, utilize the subprocess module to execute other Python processes and system commands.
[INDIRECT_PROMPT_INJECTION]: The skill fetches visual design specifications from the stylekit.top API, including aiRules and philosophy text blocks that the agent is directed to follow as constraints, creating a surface for externally sourced instructions to influence agent behavior.
[DYNAMIC_EXECUTION]: The skill employs importlib.util in scripts like eval-check.py and benchmark.py to dynamically load and execute Python modules from the local directory at runtime.

stylekit