The Agent Skills Directory

Command Execution and Configuration Modification: The skill is designed to modify repository settings, including azure-pipelines-official.yml triggers and Maestro channel associations via the darc CLI. These operations change the build and publishing behavior of the repository. To mitigate risks, the skill includes explicit instructions for the agent to present planned changes and obtain user confirmation before pushing any changes or creating pull requests.
Authentication Token Handling: The skill utilizes command substitution to fetch authentication tokens from local CLI tools (gh auth token and az account get-access-token). These tokens are passed as arguments to the darc command for authentication with Maestro and Azure DevOps. While this is a common practice in DevOps automation to avoid hardcoding secrets, handling raw tokens in command arguments is a security consideration as they may be exposed in process lists or shell history. The skill manages this by targeting these tokens toward the vendor's known infrastructure tools.
Indirect Prompt Injection Considerations: The skill processes data from external sources such as Git remotes, darc channel lists, and repository file searches (rg).
Ingestion points: Data enters the context via the output of git ls-remote, darc get-channels, gh api, and rg (SKILL.md).
Boundary markers: There are no explicit delimiters or instructions for the agent to ignore potentially malicious content within the output of these tools.
Capability inventory: The skill possesses the capability to execute shell commands, push to Git repositories, and create pull requests.
Sanitization: The skill lacks automated sanitization of tool outputs but relies on a human-in-the-loop verification process, requiring the agent to present all planned actions for user approval before execution.

setup-darc-branch-publishing