The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructs the agent on how to generate Python source code and configuration files which are then installed as editable packages into the agent's persistent Python kernel (typically at ~/.prime/agent/kernel-venv). This mechanism allows for the runtime creation and execution of new executable modules.
[INDIRECT_PROMPT_INJECTION]: The skill's core purpose is to transform user-supplied content—such as workflows, scripts, or prompts—into reusable and persistent skills. The instructions do not define sanitization or boundary markers, creating a surface where malicious instructions in the input could be converted into a persistent malicious capability.
Ingestion points: User-provided workflows, scripts, and instructions intended for conversion into new skills in SKILL.md.
Boundary markers: None specified in the instructions for the agent.
Capability inventory: The skill allows writing files to the local file system, modifying Python packages, and executing code in a persistent environment as detailed in references/python-skills.md.
Sanitization: No methods for sanitizing or escaping user-provided content are described.
[REMOTE_CODE_EXECUTION]: The skill facilitates the installation of third-party dependencies through pyproject.toml and describes methods for executing these generated capabilities using shell commands or direct kernel calls.

skill-creator