The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run shell commands to manage database tunnels and execute queries.
Evidence: lsof -nP -iTCP:3308 -sTCP:LISTEN is used to verify the state of the SSH tunnel.
Evidence: php artisan tinker --execute='...' is used to perform database operations from the command line.
[DYNAMIC_EXECUTION]: The skill utilizes runtime execution of code strings through a CLI interface.
Evidence: The use of the --execute flag with php artisan tinker allows for the execution of arbitrary PHP code within the application environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes results from an external production database, creating a surface where untrusted data enters the agent's context.
Ingestion points: Query results from DB::connection("prod_read")->select(...) as described in SKILL.md.
Boundary markers: None defined to isolate database results from the agent's instructional context.
Capability inventory: Shell command execution, database read access, and the ability to modify local scaffolding files.
Sanitization: The skill contains explicit instructions to avoid sensitive fields (names, addresses, message bodies) and to use only synthetic values when updating local scaffolding.

sociuu-baseline