The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes the alibaba/open-code-review GitHub Action. This is an official tool from Alibaba, a well-known organization. The workflow templates correctly pin the action to a specific commit SHA (1c8f930fc923753b17b80f633aea54274fc83825), which is a security best practice to prevent supply chain attacks via tag floating.
[CREDENTIALS_SAFE]: The skill follows secure practices for handling API keys. It explicitly instructs the user to pipe keys directly from a local environment file into GitHub Secrets using the gh secret set command, ensuring that sensitive tokens are never printed to logs or committed to version control.
[COMMAND_EXECUTION]: Uses legitimate command-line tools such as gh, curl, and jq for repository configuration and API health checks. The curl command in the workflows is used for a 'preflight' probe to verify LLM availability before initiating a full review run.
[PRIVILEGE_ESCALATION]: The workflow templates implement robust access controls. The automated review is gated to the repository maintainer (wakqasahmed), and the manual review trigger requires a comment from a authorized user defined in vars.MAINTAINER_USERNAME, preventing external contributors from spending the repository's LLM budget.
[INDIRECT_PROMPT_INJECTION]: While the skill processes untrusted pull request data, it mitigates risk by explicitly instructing the workflows not to check out or execute any code supplied in the PR. The code review action operates by diffing content remotely through the GitHub API, significantly reducing the attack surface for code-based injection attacks.

open-code-review-setup