The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external CI/CD configuration files (e.g., Jenkinsfile, GitLab-CI) which could potentially contain adversarial instructions targeting the conversion process. * Ingestion points: The workflow explicitly requires source files from the user in Phase 1. * Boundary markers: Instructions strictly mandate converting 'only' what is in the source and prohibit inventing workflows from descriptions. * Capability inventory: The agent has capabilities to write to the repository filesystem, create Pull Requests, and execute the actionlint tool. * Sanitization: The process requires validation of all generated YAML using actionlint and mandates pinning external actions to specific commit SHAs for integrity.
[EXTERNAL_DOWNLOADS]: Fetches and installs the actionlint utility to validate generated workflows. It also references and utilizes GitHub Marketplace actions from verified providers.
[COMMAND_EXECUTION]: Performs localized filesystem operations to archive legacy CI/CD files, including directory creation (mkdir) and moving or deleting original configuration files during the migration process.

migration-core