The Agent Skills Directory

[COMMAND_EXECUTION]: The skill constructs and executes gh pr create commands. It mitigates command injection risks by utilizing quoted heredocs (cat <<'EOF'), which ensures that PR body content is treated as a literal string by the shell.
[SAFE]: A dedicated safety-checks.md file defines a Self-Refine evaluation loop. The agent is required to check its own command output for proper quoting and special character handling before execution, providing a layer of protection against unexpected input.
[SAFE]: The skill only accesses non-sensitive repository metadata like branch names and commit logs, which are necessary for PR metadata generation.

pr-creation