The Agent Skills Directory

[COMMAND_EXECUTION]: The provided Kubernetes deployment templates in SKILL.md use the --insecure-skip-tls-verify=true flag. This configuration disables SSL/TLS certificate validation, exposing the CI runner to potential man-in-the-middle attacks when communicating with the Kubernetes API server.
[DYNAMIC_EXECUTION]: The skill includes a pattern for 'Dynamic Child Pipelines' where a Python script is used to generate a YAML configuration file at runtime, which is then executed. This practice involves dynamic code generation and execution within the pipeline environment.
[INDIRECT_PROMPT_INJECTION]: The dynamic pipeline generation pattern creates a vulnerability surface for indirect prompt injection. If the generation script (generate_pipeline.py) processes untrusted data from repository files or external sources without sanitization, malicious instructions could be injected into the child-pipeline.yml and executed by the runner.
Ingestion points: The generate_pipeline.py script in the generate-pipeline job acts as the entry point for data that defines the child pipeline structure.
Boundary markers: The templates do not include boundary markers or instructions to isolate or ignore untrusted content during pipeline generation.
Capability inventory: The skill utilizes high-privilege tools such as kubectl for deployment, docker for registry operations, and terraform for infrastructure management.
Sanitization: There are no explicit sanitization or validation logic shown for the inputs processed by the pipeline generation script.

gitlab-ci-patterns