The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides templates for processing untrusted data from GitHub event payloads, which is a known attack surface for indirect injection if not handled correctly.
Ingestion points: Pull Request titles and other event data are accessed in references/external-files.md and references/security.md.
Capability inventory: The skill utilizes the GitHub API (github.rest) and shell command execution (exec.exec) within its scripts and examples.
Boundary markers and Sanitization: The skill includes a dedicated security guide (references/security.md) that explicitly mandates the use of environment boundaries (process.env) and validation to prevent injection, mitigating the risk.
[DYNAMIC_EXECUTION]: The skill uses await import() to load local scripts from the repository workspace and the exec helper to execute shell commands, such as checking version information with the uv tool.
Evidence: assets/check-stubs-version-job.yml imports local modules using the ${process.env.GITHUB_WORKSPACE} path.
Evidence: assets/scripts/version-check-shared.mjs uses exec.exec('uv', ...) to interact with the runtime environment.
[EXTERNAL_DOWNLOADS]: The assets/package.json file references the official actions/github-script repository and other standard developer tools. The skill also references external logic from the author's own repository (kjanat/kp2bw). These are legitimate resources for the skill's intended purpose.
Evidence: assets/package.json includes "@actions/github-script": "github:actions/github-script".
Evidence: references/external-files.md links to https://github.com/kjanat/kp2bw/tree/master/scripts for template logic.

github-script