The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts scripts/dump_dom_and_bindings.py and scripts/generate_pilot_tests.py perform dynamic code loading and execution. They use importlib and exec_module to load Python modules from computed paths within the project directory to inspect application classes and generate metadata for the agent.
[COMMAND_EXECUTION]: The validation script scripts/self_check.py executes shell commands via subprocess.run to test the skill's internal scaffolding functionality.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes content from local project files that may contain instructions designed to influence agent behavior.
Ingestion points: Python and TCSS files within the project root are read and processed by auditing scripts such as scripts/inspect_textual_project.py and scripts/audit_textual_project.py.
Boundary markers: Analyzed content is passed to the agent without explicit boundary markers or instructions to ignore embedded instructions.
Capability inventory: The skill possesses the capability to write files to the local filesystem and execute local scripts.
Sanitization: The skill uses ast.parse for structural code analysis but does not sanitize or escape the content of files before presentation to the agent.

textual-tui