The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes infrastructure management commands including kubectl for cluster status checks, curl for Airflow API connectivity tests, and make sync-dag for deploying configuration and code to S3. These are standard operations for its stated purpose of DAG orchestration.
[DYNAMIC_EXECUTION]: The skill dynamically generates Python scripts (.py) and Kubernetes manifests (.yaml) based on existing reference templates and user requirements. It performs validation using python -c to ensure the generated files are syntactically correct before deployment.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes user-provided pipeline descriptions to generate executable code and prompts.
Ingestion points: User-supplied pipeline stage sequences and domain-specific prompt requirements (SKILL.md, Step 1).
Boundary markers: Not explicitly defined in the generated code; the skill relies on the agent following reference guides (prompt-authoring.md) from component skills.
Capability inventory: File system writes, kubectl cluster access, Airflow API interaction, and S3 deployment via make commands (SKILL.md, Step 8).
Sanitization: The skill mitigates risks by requiring the agent to map user input to a predefined list of shared TaskGroup classes and adapt existing callables rather than writing logic from scratch.
[CREDENTIALS_UNSAFE]: The skill requests a file path for cluster credentials to facilitate kubectl operations. This requirement is disclosed in the skill metadata, and the instructions explicitly warn against logging credential file contents.

write-dag