plan

Creates an approved plan.md by interviewing requirements, measuring impact scope, and decomposing work into verifiable tasks.

3|Updated May 20, 2026
One-click install
npx skills add https://github.com/jongcheol-pak/claude-harness-pjc --skill plan-jongcheol-pak
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: plan
Source: https://github.com/jongcheol-pak/claude-harness-pjc/tree/main/plugins/pjc/skills/plan
Command: npx skills add https://github.com/jongcheol-pak/claude-harness-pjc --skill plan-jongcheol-pak

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Code changes made without a plan lead to missed files, broken callers, and unverified "done" reports. This Skill forces a structured planning phase before any code is touched: requirements are confirmed through a one-question-at-a-time interview, the impact scope is measured with real commands, and the work is split into tasks that each carry files, acceptance criteria, and verification commands. ## Core Features & Use Cases - Requirements Interview: Confirms intent through single-question rounds with GUESS hypotheses until confidence reaches 95%, then restates the goal for explicit approval. - Command-Verified Impact Analysis: Every factual claim in the plan must come from an executed command recorded in an Investigation Log table. - Task Decomposition with Review: Splits work into T1/T2 tasks with Files, acceptance criteria, and verification commands, then runs a plan-reviewer subagent before user approval. - Use Case: When asked to "add a password-reset link and refactor the login flow," the Skill interviews the user, greps the codebase for affected callers, writes plan.md with verifiable tasks, gets approval, and hands off to the implement skill. ## Quick Start Ask the assistant to plan a multi-file code change, such as "plan adding a password reset feature to the login screen," and it will interview you, measure the impact scope, and produce an approvable plan.md.

Frequently Asked Questions about plan

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

FAQPage Schema
How do I plan a code change before implementing it with Claude Code?▼

Invoke the plan skill with your change request. It interviews you one question at a time until requirements are 95% confirmed, measures the impact scope with real commands, and writes a plan.md with tasks, acceptance criteria, and verification commands for your approval.

When does the plan skill trigger versus editing code directly?▼

It triggers for multi-item requests, multi-file changes, logic or signature changes, and new functions or classes. It is skipped only for a single trivial change such as a typo, label, config value, or simple value substitution of three lines or fewer.

Does the plan skill work with any programming language or project?▼

Yes, it is language-agnostic. It reads build and test commands from the project's AGENTS.md file, creating a minimal one if missing, so any project with known build and test commands can be planned and verified.

What happens after the plan is approved?▼

After approval, the skill commits the intent file recording the approved requirements, then hands off in the same turn to the pjc:implement skill, which executes the approved plan tasks through completion.

Why does the plan require an Investigation Log with executed commands?▼

Every factual claim in the plan must come from an actually executed command recorded in a three-column table. Claims without a command entry are treated as guesses and cannot be included, preventing plans built on assumptions about the codebase.