Task

New Task

Compute the task directory path, then run the script. If no argument is provided, use the current conversation to fill in the details.

1. task-name: hyphenated, lowercase, 2-4 words summarizing the argument (e.g., buggy-auth).
2. issue-number: deduce from argument or branch name (e.g., ECMS-12345). Omit if not applicable.
3. task-title: title-cased task-name, prefixed with issue-number if applicable (e.g., ECMS-12345: Buggy Auth).
4. project-dirname: basename of the current project directory (e.g., my-project).
5. folder: {YYYY-MM-DD}_{project-dirname}_{issue-number}-{task-name} (omit {issue-number}- if none).
6. task-dir: ~/.pi/plans/{folder}.

Run:

~/.pi/agent/skills/task/task.sh "<task-dir>" "<task-title>"

The script creates the directory and 1-TASK.md, then outputs the path.

Once created:

1. Read and present the 1-TASK.md contents.
2. Ask the user how they want to proceed. Recommend an option based on the details provided and task complexity:
   • Research web
   • Brainstorm design
   • Add more detail
   • Implement directly
3. Handle the choice:
   • Brainstorm design → Use the Pi-adapted brainstorming skill, using 1-TASK.md as input and writing 2-DESIGN.md in the same task folder.
   • Add more detail → Accept more input, merge it into 1-TASK.md, then ask again how to proceed.
   • Implement directly → Implement directly only if the task is small, clear, and does not require design.
   • Research web → Ask for the research topic or sources, gather findings if web access is available, update 1-TASK.md, then ask again how to proceed.

Prefer the full workflow for non-trivial changes:

1-TASK.md → Pi-adapted brainstorming → 2-DESIGN.md → Pi-adapted writing-plans → 3-PLAN.md → Pi-adapted executing-plans or subagent-driven-development.