The Agent Skills Directory

[SAFE]: The skill contains comprehensive documentation and instructions (SKILL.md and references/) designed to enforce high security standards for GitHub infrastructure. It explicitly warns against dangerous patterns like pull_request_target abuse and encourages the use of security scanners like actionlint and zizmor.
[INDIRECT_PROMPT_INJECTION]: The skill includes an evaluation suite (evals/) that provides examples of vulnerable code for the agent to fix. These inputs are clearly delimited by boundary markers (e.g., =============== FILE: ... ===============), and the skill's core instructions mandate the sanitization of all inputs and outputs, effectively mitigating risks associated with processing untrusted content during repository setup.
[EXTERNAL_DOWNLOADS]: The skill references established GitHub Actions and reference implementations from the author's organization (uinaf). These are documented neutrally as trusted vendor resources for repository automation and do not involve the execution of unverified remote code.
[CREDENTIALS_UNSAFE]: While the skill discusses secret management, it correctly avoids hardcoding credentials, instead using placeholders like RELEASE_APP_PRIVATE_KEY and directing users toward secure practices such as GitHub Environments and runtime secret stores.

gh-setup