The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands for git and the GitHub CLI (gh) to perform repository management tasks.
[DATA_EXFILTRATION]: The skill transmits local code changes and branch content to a remote GitHub repository as part of its core functionality.
[COMMAND_EXECUTION]: Provides functionality to bypass GitHub branch protection rules via the --admin flag with the gh tool, though this is explicitly gated by instructions to seek user confirmation and verify repository ownership.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface due to its analysis of repository data.
Ingestion points: Untrusted data enters the agent context through the analysis of repository history via git log and current changes via git diff in SKILL.md.
Boundary markers: No explicit delimiters are specified for separating ingested content from instructions.
Capability inventory: The skill can perform file commits, remote pushes, and PR creation/merging across multiple scripts.
Sanitization: Employs mandatory user confirmation checkpoints before any destructive or outward-facing action (committing, pushing, and merging) to ensure human oversight.

raise-pr