jeninh

creating-tasks-through-lunatask

0
0
# Install this skill:
npx skills add jeninh/ampskills-dotfile --skill "creating-tasks-through-lunatask"

Install specific skill from multi-skill repository

# Description

Creates tasks in Lunatask for todos, reminders, deferred work, and handoffs. Use when the user wants to capture something for later, whether for themselves or for a future agent session.

# SKILL.md


name: creating-tasks-through-lunatask
description: Creates tasks in Lunatask for todos, reminders, deferred work, and handoffs. Use when the user wants to capture something for later, whether for themselves or for a future agent session.
compatibility: Requires Lunatask MCP or CLI tools
license: AGPL-3.0-or-later
metadata:
author: Amolith [email protected]


Tasks: Reminders, todos, bug reports, feature ideas. Self-contained; the note captures what and why for the user.

Handoffs: Work a future agent will pick up in a fresh context window. Requires careful framing so the receiving agent knows where to start and what we learned. See handoff.md before creating these.

If unclear which type, ask the user "is this a task or a handoff?"

Tone

Always include a warm, friendly, personal, and conversational note from you to the user. Address them by name if you know it. Otherwise, just start with a friendly greeting omitting their name. Task notes are exclusively a friendly note from you to the user, while handoffs also include a thorough prompt for another agent to pick up. See handoff.md#tone for more detail.

Formatting

Title

Titles must be clear:

  • Useful and concise
  • Conversational
  • Include the project name only if goal doesn't exactly match the project name.
  • A task for the go-lunatask client library in area projects and goal lunatask would require go-lunatask in the title.
  • A task for crush in area projects and goal crush should not include crush in the title.

Examples:

  • Accept 'completed' as status alias in lune
  • Implement ACP event sink in Crush
  • Investigate why lune habit track silently fails

Status

New tasks usually get status later. When beginning work on a task, update to in-progress, then completed when done (or use done shortcut).

Note structure

Tasks:

> [Friendly greeting], [context if helpful]
>
> [What needs doing and why]
>
> [Optional: relevant links, files, or commands if they'd help]

Handoffs:

> [Friendly greeting], [context if helpful]
>
> [What needs doing and why]

---

[See references/handoff.md for how to write this section]

References

Include relevant links when they'd helpโ€”issues, commits, docs, files. For handoffs, gather these proactively from previous messages; for tasks, include what's useful.

Appending to existing notes

Only append; never overwrite unless explicitly asked.

Lunatask only supports full replacement, so reproduce the original note verbatim, then add:

---

[New content here]

Surfacing tech debt

While working, you may notice code smells, outdated patterns, missing tests, or other improvement opportunities related or unrelated to the current task. We might also intentionally introduce tech debt, aware of the trade-offs. Don't let these observations or introductions disappear with the session; ask the user whether you should mention them.

If the user confirms and it's/they're

  • relevant to the handoff: mention briefly in context section so the receiving agent is aware
  • worth tracking separately: suggest capturing as standalone task(s) so the user can pick up or hand off to another agent later. Even as standalone tasks, make sure to include enough context that another agent could pick it up.

Working around a Lunatask bug

At the very very end of task notes or note contents, include [editor_v2]:: on its own line.


If lunatask or lune MCP tools are unavailable, see cli.md. If the CLI is also unavailable, see installing-lune.md.

# Supported AI Coding Agents

This skill is compatible with the SKILL.md standard and works with all major AI coding agents:

Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.