skills.rest
infinition
infinitionCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

long-running-work

Track multi-step work using todo lists, plan mode, kanban boards, and scheduled missions.

33|1|Updated Mar 4, 2026
Search Tags:#todo list#workflow management#task tracking#kanban#missions#plan mode#cron scheduling
One-click install
npx skills add https://github.com/infinition/LaRuche --skill long-running-work-infinition
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: long-running-work
Source: https://github.com/infinition/LaRuche/tree/main/laruche/skills/long-running-work
Command: npx skills add https://github.com/infinition/LaRuche --skill long-running-work-infinition

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Multi-step work gets lost between conversation turns: steps are forgotten after long tool outputs, plans get overwritten, and recurring objectives are never tracked. This Skill provides four mechanisms with distinct lifespans so work is tracked at the right scope instead of being dropped or wrapped in unnecessary ceremony. ## Core Features & Use Cases - Todo lists: Track steps within the current mission, marking one item in progress at a time so nothing is forgotten mid-task. - Plan mode: Create a plan.md with Context, Proposed Steps, and Approval Required sections before making structural changes, with existing plans safely renamed rather than overwritten. - Kanban board: Persist work items across sessions with dependency tracking, where completing a task automatically unblocks dependent tasks. - Missions: Define objectives pursued across many iterations, optionally on a cron schedule, with honest completion reporting via task_complete. - Use Case: A user asks for a multi-file refactor. You create a plan for approval, track execution steps with todo, and record the outcome on the user's kanban board for cross-session visibility. ## Quick Start Break my request into tracked steps with a todo list, and if the change is structural, draft a plan for my approval before editing anything.

Frequently Asked Questions about long-running-work

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I track multi-step tasks in an AI agent workflow?▼

Use a todo list with one entry per step, marking exactly one item in progress at a time and updating it as work completes. Add newly discovered steps immediately rather than holding them, since a stale todo list reports progress that did not happen.

What is the difference between a todo list and a kanban board?▼

A todo list is scratch space scoped to the current mission and discarded after. A kanban board persists across sessions, belongs to the user, and supports dependencies where completing a task unblocks the tasks waiting on it.

When should I use plan mode before making changes?▼

Use plan mode for structural changes such as multi-file edits, migrations, or deletions where the user would want veto power. Do not use it for a fix the user explicitly requested, since that only delays an already-approved action.

How do I schedule a recurring automated mission?▼

Create a mission with an objective and a cron expression such as "0 9 * * *" for daily at 09:00. Write the objective so it is executable and verifiable on its own, and confirm the cadence with the user since a scheduled mission is a standing commitment.

Why does kanban_next keep returning the same task?▼

kanban_next only reads the board and never advances anything. You must call kanban_complete with the task's UUID and a result describing what changed, which unblocks dependent tasks and lets the next call return new work.

When should I use a watcher instead of a scheduled mission?▼

Use a watcher when a condition decides rather than you, such as "warn me when X happens". A scheduled mission that repeatedly checks whether something occurred is the expensive way to write a watcher, since watcher rules cost nothing while idle.

Related Skills

Explore More Product & Management Skills→
dudqks0319-cpududqks0319-cpu

Prompt Architect

Converts vague user requests into structured development prompts through scenario options and refinement questions.

Community
Intermediate
jbactadjbactad
1

debrief

Produces standardized completion summaries and archives mission state for audits.

Community
Advanced
KanuToCLKanuToCL

condition-based-waiting

Poll for state changes instead of arbitrary timeouts in automated tests.

Community
Intermediate
coco-researchcoco-research
264

gsd-fast

Execute trivial code changes directly in the current context.

Community
Basic