The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill identifies project files, SQL dumps, and migration comments as untrusted data that could potentially contain malicious instructions. * Ingestion points: The agent reads 'composer.json', 'phinx.php', and migration files in 'db/migrations/' using shell commands. * Capability inventory: The skill utilizes subprocess calls to execute 'vendor/bin/phinx' and a local Python validation script. * Boundary markers: Includes a 'Prompt injection guard' section that explicitly instructs the agent to treat repository configuration as hostile and disregard instructions within data. * Sanitization: Employs 'scripts/validate_migration.py' to detect hardcoded credentials and placeholders in the migration files.
[COMMAND_EXECUTION]: The skill uses shell commands to inspect the project structure and manage database migrations. * Evidence: Execution of 'cat', 'ls', and 'vendor/bin/phinx' for status checks, migrations, and rollbacks.
[DYNAMIC_EXECUTION]: The skill generates PHP migration files based on user requirements and executes a local Python validation script. * Evidence: Gate 4 involves creating PHP files (via 'vendor/bin/phinx create'), and Gate 5 executes 'python3' to run the 'validate_migration.py' script.

migrations-phinx