The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses git and the GitHub CLI (gh) to view, edit, update, and merge pull requests. It also uses git diff to analyze changes between environments and gh repo view to identify branch configurations.
[REMOTE_CODE_EXECUTION]: The skill is designed to identify and execute migration commands defined within the target repository's project files (e.g., in package.json scripts, README files, or migration tool configurations). Since these commands are sourced from untrusted repository content and executed with production-level authorization, this represents a significant remote code execution vector.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository to determine its release and promotion logic, creating a broad attack surface for indirect instructions.
Ingestion points: The agent is instructed to read the repository's README, 'agent instructions', CI/deploy workflows, and Pull Request metadata to determine how to proceed with deployments.
Boundary markers: No delimiters or safety warnings are provided to prevent the agent from obeying malicious instructions embedded within these external files.
Capability inventory: Across its scripts, the agent can merge code into main and environment branches, run arbitrary shell commands for migrations, and modify pull request targets.
Sanitization: There is no mechanism described for validating or escaping the instructions or commands retrieved from the repository files before they are processed or executed.
[PROMPT_INJECTION]: The skill includes instructions that explicitly override standard safety protocols: 'The user's request authorizes every merge and every in-scope migration, including production, so work without asking for approval'. This directive encourages the agent to bypass human-in-the-loop verification for sensitive production operations.

merge-and-migrate