The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local system commands including git, grep, and python3. These tools are used for inspecting repository state (e.g., retrieving remote URLs and commit history) and for validating the syntax of generated YAML configuration files. These operations are restricted to the local environment and the project's scope.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data by reading local repository files such as lockfiles, manifests, and existing configuration to infer project settings. This creates a surface for indirect prompt injection if the project files contain adversarial content. However, the risk is mitigated by the skill's instruction to present all planned changes and wait for user approval before writing any files.

github-essentials