The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/execute_action.py uses subprocess.run to execute other scripts within the skill. It utilizes a hardcoded list of valid scripts and passes arguments as a list, minimizing the risk of arbitrary command injection.
[PROMPT_INJECTION]: The skill contains a directive to the agent to execute actions 'outside of the sandbox'. This is a request to bypass environment security restrictions, typically to allow the network access required for the Asana API.
[PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes untrusted data from the Asana API.
Ingestion points: Content such as task names, notes, and comments are fetched from Asana in scripts/get_task.py, scripts/list_tasks.py, scripts/search_tasks.py, and scripts/my_tasks.py.
Boundary markers: There are no markers or instructions provided to the agent to prevent it from following commands that might be embedded in task data retrieved from external sources.
Capability inventory: The skill allows for writing data back to Asana (create_task, update_task, add_comment) and executing local Python scripts through the dispatcher.
Sanitization: The skill uses argparse for command-line arguments but does not implement sanitization or escaping for the data retrieved from the API before it is presented to the agent's context.

asana