<discovery>
The library reference shows abbreviated descriptions. When the summary isn't enough to confidently pick the right task or workflow, inspect the source files directly using run_shell before deciding.

Tasks directory: $tasks_path
- List all tasks:       run_shell(command="ls $tasks_path")
- Read a task file:     run_shell(command="cat $tasks_path/<name>.py")
  Key class attributes: `tags`, `description`, `long_description`, `input_types`, `output_types`, `opts`

Workflows directory: $workflows_path
- List all workflows:   run_shell(command="ls $workflows_path")
- Read a workflow file: run_shell(command="cat $workflows_path/<name>.yaml")
  Key YAML fields:      `description`, `long_description`, `tags`, `input_types`, `tasks`

Profiles directory: $profiles_path
- List all profiles:    run_shell(command="ls $profiles_path")
- Read a profile file:  run_shell(command="cat $profiles_path/<name>.yaml")
  Key YAML fields:      `description`, `category`, `opts` (the option overrides the profile applies)

Use this when:
- The reference description is too brief to judge fit between multiple candidates
- You need to verify exact supported options or input/output types before running
- You want to understand which sub-tasks a workflow runs before committing to it
- You need to know what options a profile sets before applying it
</discovery>
