make-plan

SkillProductivity

Lets your agent create a detailed, phased implementation plan before starting a task.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the make-plan skill

About this capability

Break the chosen proposal into ordered issues/tasks with dependencies. Use after a proposal is chosen, before implementing.

What this skill tells your AI

The instructions your AI receives, as published by nilswidal/loobster in .agents/skills/make-plan/SKILL.md and read by ahel’s review.

Break the chosen proposal into a concrete implementation plan.

Behavior

  • Do NOT write any code. Only create the plan.
  • Clarify the code change scope, constraints, and ordering before creating anything.
  • Use templates/design-doc-template.md as a reference for what sections to cover.

Steps

  1. Read the relevant research and proposal documents.
  2. If Linear MCP tools are available:
    • Create a parent issue that serves as the epic/tracker for the full plan. Include context, requirements, design decisions, and any security/testing notes in the description.
    • Break into ordered sub-issues under the parent, each small enough for a single commit. Include affected files, acceptance criteria, and blockers.
    • Set blocking relationships where steps depend on each other.
    • Mark independence explicitly. For each sub-issue, state whether it is independent (no blocking edge, touches a disjoint set of files) or dependent. This is what lets the Implement phase run independent slices in parallel (see run.md Phase 4). Two sub-issues that edit the same file are not independent — give them a blocking edge to force serialization.
  3. If Linear is not available:
    • Create a plans/ directory with a parent plan .md file and individual task .md files.
  4. Regardless of whether Linear is available, also create Claude Code Tasks using TaskCreate for each sub-task. This provides persistent task tracking that survives across sessions.
    • When invoked under a goal-loop (/loop): tag each Task's metadata with the active goalId and a RICE score per .agents/skills/backlog-scoring/SKILL.md, so the sub-tasks join the prioritized backlog.
  5. Present the plan to the user for review.

Notes

  • Keep solutions simple and focused. Only plan changes that are directly requested.
  • Each sub-issue should map to a vertical slice when possible (DB + API + UI together).

Signals

GitHub stars
36
Forks
2
Last commit
Aug 2026
Advanced
Catalog kind
skill
Gateway key
make-plan
Source
github.com/nilswidal/loobster