The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The plugin development standards in references/plugin.md describe a mechanism for dynamic code injection. Both App-level and Extend-level plugins allow custom logic, API routes, and service layers to be loaded and executed by the core system at runtime.\n- [EXTERNAL_DOWNLOADS]: The framework architecture supports plugins with their own requirements.txt files, suggesting a process for downloading and installing unverified external Python dependencies during plugin setup.\n- [COMMAND_EXECUTION]: The documentation frequently references the use of an fba CLI tool for essential development tasks, including database migrations (fba alembic) and code formatting (fba format), which involves executing system-level commands.\n- [PROMPT_INJECTION]: The skill architecture presents an indirect prompt injection surface. (1) Ingestion points: The skill processes external plugin configuration files (plugin.toml) and requirements.txt files. (2) Boundary markers: No delimiters or instructions to ignore embedded malicious content are mentioned in the guides. (3) Capability inventory: The skill utilizes dynamic code loading and CLI command execution. (4) Sanitization: The documentation lacks specifications for sanitizing or validating data provided by external plugins before it affects system behavior.

fba