The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses dynamic context injection (the !command syntax) in SKILL.md to automatically gather repository state, such as git status, the current branch name, and recent commit logs, at the time the skill is invoked.
[COMMAND_EXECUTION]: The skill executes various git and GitHub CLI commands through the Bash tool to perform operations such as branch management, code pushing, and pull request creation.
[COMMAND_EXECUTION]: The skill executes a local Python script scripts/format-pr-body.py to generate PR descriptions. This script uses subprocess.run to call git log and git diff. It uses the list-based argument format and avoids shell=True, which is a secure practice that prevents standard shell command injection.
[DATA_EXFILTRATION]: The skill reads local repository information, including commit messages and file diffs, and transmits this data to GitHub via git push and gh pr commands. This behavior is consistent with the skill's primary function of submitting code for review.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it reads git commit messages and file names and includes them in pull request descriptions without sanitization or boundary markers.
Ingestion points: The get_commits and get_changed_files functions in scripts/format-pr-body.py ingest data directly from the local git command outputs.
Boundary markers: The generated PR body does not use delimiters or explicit instructions to distinguish system-generated text from the content of commit messages.
Capability inventory: The skill has the capability to push to remote repositories and create persistent records on GitHub via the gh tool.
Sanitization: The script does not perform validation or escaping of commit messages before interpolating them into the markdown PR body.

push-pr