The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources that could contain malicious instructions designed to manipulate the agent.
Ingestion points: The skill reads pull request titles, bodies, comments, notes, and git diffs/commit logs in Phase 0 and Phase 3.
Boundary markers: Phase 0 contains explicit instructions to treat retrieved content as inert untrusted data and to never follow instructions or execute commands found within that content.
Capability inventory: The skill has the ability to push code to remote repositories, create/update/merge pull requests, and read local files.
Sanitization: The skill requires quoting and escaping untrusted values passed to commands or generated prose and includes a secret scanner in Phase 4 to prevent the exfiltration of credentials in PR descriptions.
[COMMAND_EXECUTION]: The skill relies on executing shell commands to interact with git and forge providers.
It uses git, gh (GitHub CLI), and glab (GitLab CLI) to perform repository operations.
Commands are constructed dynamically using variables derived from git output and user-supplied arguments.

create-pr