The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs shell command execution to manage the migration workflow.\n
Evidence: The script scripts/migrate.py uses subprocess.run to execute various Git commands, including git rev-parse, git checkout, git add, and git commit to handle branch management and versioning of the changes.\n- [CREDENTIALS_UNSAFE]: The skill interacts with sensitive configuration files that are known to contain credentials.\n
Evidence: scripts/migrate_config.py is designed to read and modify .env files, which are standard locations for storing sensitive secrets, API keys, and other private configuration data.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted source code and possesses significant capabilities.\n
Ingestion points: scripts/migrate.py and its migration modules ingest the entire content of a user-provided repository, including Python source files, requirements lists, and configuration manifests.\n
Boundary markers: No boundary markers or 'ignore embedded instructions' warnings are implemented to protect the agent while it processes external repository content.\n
Capability inventory: The skill has the authority to overwrite any file within the provided repository path and can execute Git commands via subprocess.\n
Sanitization: The code transformation logic in scripts/migrate_routes.py and scripts/migrate_tests.py relies on simple regular expressions. This method is not robust and could be exploited by maliciously crafted input strings in the target codebase to inject unwanted or malicious code during the transformation process.

framework-migration-assistant