The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill clones the repository https://github.com/lutzseverino/repository-standards.git to a temporary directory to retrieve release-specific instructions. This repository belongs to the skill author.
[COMMAND_EXECUTION]: The scripts/select-release script executes git via subprocess to clone the repository and verify the integrity of the local checkout, including checks for exact tag matching and workspace cleanliness.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and follow instructions from an external source downloaded at runtime.
Ingestion points: The agent is instructed to read and follow the SKILL.md file found within the cloned repository at the path .agents/skills/create-repository/SKILL.md.
Boundary markers: The instructions do not employ delimiters or specific safety directives to ensure the agent ignores potentially adversarial instructions embedded in the downloaded content.
Capability inventory: The instructions from the external source are expected to facilitate repository and GitHub mutations.
Sanitization: While the script performs strong integrity checks on the checkout (git tag verification and dirty-tree detection), it does not validate the natural language content of the instructions against potential prompt injection attacks.

create-repository