The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes repository files such as manifests and Dockerfiles to determine CI requirements. This ingestion of untrusted data represents a potential attack surface, but the skill mitigates risk by requiring the use of documented project commands and explicitly forbidding the reading of credentials or environment variables.
[DATA_EXPOSURE]: The skill contains robust safety guidelines that prohibit the agent from accessing sensitive files like .env, kubeconfigs, or credential stores. It mandates the use of sanitized placeholders for secrets in all generated configuration files.
[COMMAND_EXECUTION]: The skill involves identifying build and test commands within a repository to wire them into a CI pipeline. While it interacts with project commands, it does so within the scope of repository configuration and explicitly requires user validation of the generated output.

setup-ci