task-scoper

Skill: Task Scoper

Purpose

Prevents scope creep by explicitly defining the boundaries of a task before implementation begins. Ensures the agent works on exactly what was asked — no more, no less.

When to Use

• At the start of any non-trivial implementation task
• When an issue or instruction is broad or ambiguous
• When tempted to "clean up" or "improve" things unrelated to the task
• Before invoking implementation-planner or parallel-agents

Process

1. Read the Task Carefully

• Identify the single core ask
• List explicit requirements stated in the task
• List implicit requirements (things obviously needed but not stated)
• List things explicitly excluded (if any)

2. Define the Boundary

Write a one-paragraph scope statement:

"This task will [do X] by [method Y]. It will NOT [do Z] even if Z seems related."

3. Identify Temptations

• List things you might be tempted to change that are out of scope
• Examples: refactoring unrelated code, updating docs not touched by the change, fixing unrelated bugs noticed along the way

4. Lock the Scope

• Commit to the scope statement before writing any code
• If during implementation you discover the scope was wrong, stop and re-scope rather than expanding silently

5. Out-of-Scope Findings

• Log any out-of-scope issues found during implementation as separate notes
• These can become new issues or tasks — do not fold them into the current task
• Use learn-rule if the out-of-scope finding is a pattern worth capturing

Output

• A clear scope statement (can be inline reasoning)
• A list of intentionally excluded items
• A focused implementation that respects the defined boundaries

Notes

• Scope creep is one of the most common failure modes for AI agents
• A smaller, correctly-scoped change is always better than a larger, drifted one
• When in doubt, do less and document what was left out