The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Node.js scripts to execute the git and gh (GitHub CLI) tools. This is necessary for its primary function of inspecting repositories, committing changes, and managing pull requests. The commands are executed with structured arguments rather than raw shell strings, which reduces the risk of command injection.
[DATA_EXFILTRATION]: The skill processes local repository data and configuration files. It specifically includes a security safeguard in scripts/core.mjs to detect and block the inclusion of local system paths (e.g., /Users/, /home/, C:\) in the generated pull request body, preventing accidental disclosure of local environment details.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources such as GitHub issue bodies, pull request templates, and repository-specific instructions. While this creates a surface for indirect prompt injection, the documentation explicitly instructs the agent to treat this output as data rather than instructions. The proseLines utility in the script also attempts to ignore content within code fences to mitigate accidental execution of embedded patterns.
[SAFE]: The skill implements a robust journaling and reconciliation system to handle network failures or interrupted operations safely. It ensures that pull requests are not duplicated and that the state of the repository and remote are verified before any write operations occur.

create-pr