The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes several command-line interfaces for its core functionality:
Git/GitHub CLI: Uses git for branch management and pushing, and gh for creating, viewing, and editing pull requests.
Bun: Executes bun run build, bun run stylecheck, and bunx oxfmt to format code and verify the build status.
Vercel CLI: Conditionally uses vercel inspect to resolve deployment URLs if the CLI is authenticated.
[INDIRECT_PROMPT_INJECTION]: The skill implements a mechanism to poll and parse external data (GitHub PR comments) to extract Vercel preview links.
Ingestion points: Polls PR comments for a specific Vercel project row to retrieve deployment links.
Boundary markers: No specific delimiters are used for the parsed text, but the skill scope is limited to appending links to a specific section.
Capability inventory: Includes the ability to execute shell commands (gh, git, bun, vercel), write to the system temporary directory, and modify PR metadata.
Sanitization: The skill follows a critical security best practice by writing the parsed/updated PR body to a temporary Markdown file and using the --body-file flag, which effectively prevents command injection that could occur if the data were passed as an inline shell argument.

pr