The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted Python source code (Prefect flows) provided by the user to perform its primary function. This ingestion creates a surface where malicious instructions embedded in the code could potentially influence the agent's logic during the analysis or code generation phases.
Ingestion points: Phase 1 of the workflow in SKILL.md requires reading user-provided flow files, deployment configurations, and supporting modules.
Boundary markers: The skill does not explicitly define delimiters or instructions for the agent to ignore instructions embedded within the processed data.
Capability inventory: The skill performs file system operations (reading files, executing zenml init), generates executable Python code, and suggests system-level tool configuration (claude mcp add).
Sanitization: No explicit sanitization or validation of the ingested code content is described.
[EXTERNAL_DOWNLOADS]: The skill suggests the installation of an MCP server from the vendor's official documentation domain at docs.zenml.io. This is a legitimate functional reference to a vendor-owned resource.
[COMMAND_EXECUTION]: The skill instructs the agent to suggest or execute commands such as zenml init to initialize projects and claude mcp add to configure the environment. These are standard operations within the ZenML ecosystem and follow vendor-recommended practices.

prefect-to-zenml-migration