The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the GitHub CLI (gh) and Git to automate pull request workflows. It reads branch names, commit logs, and diffs to generate content.
[REMOTE_CODE_EXECUTION]: The skill demonstrates high security awareness by explicitly mitigating shell injection risks when calling external tools. In Step 3, it instructs the agent to write the PR title to a temporary file and use a quoted heredoc (cat <<'EOF') for the PR body. This ensures that potentially malicious content within commit messages (such as backticks or shell variables) is not executed by the host shell during the pull request creation process.
[INDIRECT_PROMPT_INJECTION]: The skill identifies commit messages and diffs as untrusted data sources that may contain adversarial instructions.
Ingestion points: Data enters the context via git log and git diff commands in SKILL.md.
Boundary markers: The instructions explicitly mandate the use of clear boundary markers (e.g., --- BEGIN GIT LOG ---) to isolate untrusted data from the agent's instructions.
Capability inventory: The agent has the ability to run shell commands (gh pr create) and write files.
Sanitization: The skill implements sanitization through shell-safe command construction (temp files and quoted heredocs) as noted in the RCE section.

pr-branch