plan

Creates grounded, phased implementation plans verified against the actual codebase.

Updated May 13, 2026
One-click install
npx skills add https://github.com/andrewcodesit/skills --skill plan-andrewcodesit
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: plan
Source: https://github.com/andrewcodesit/skills/tree/main/skills/project-management/plan
Command: npx skills add https://github.com/andrewcodesit/skills --skill plan-andrewcodesit

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Agents often write plans that are ignored, guessed at, or drift from reality during implementation. This Skill forces a real implementation plan to exist before any code is touched, grounding every claim in the actual repository and gating execution phase by phase. ## Core Features & Use Cases - Codebase-grounded planning: Reads AGENTS.md, context maps, and the actual implementation so every file path and API in the plan is verified, not recalled. - Phased plans with verification gates: Splits larger work into ordered phases, each ending in a gate that must pass before the next phase starts, with a Task Ownership table for concurrency and risk classification. - Structured user questioning: Uses a strict question format (references/question-format.md) to resolve real unknowns before writing, recording all answers in the plan's Decisions section. - Use Case: Before implementing a new API endpoint, invoke the skill to inspect the repo, grill you on open design questions, and produce a plan saved to ~/.agents/plans/<repo>/ with contracts, pre-mortem scenarios, and validation steps. ## Quick Start Ask the agent to plan the feature you want to build, for example by saying "write a plan for adding user authentication to the settings page".

Frequently Asked Questions about plan

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

FAQPage Schema
How do I write an implementation plan before coding with an AI agent?▼

Invoke the plan skill with a phrase like "plan X" or "write a plan for X". It inspects the repository, asks structured questions about real unknowns, then saves a phased plan with verification gates to ~/.agents/plans/<repo-name>/ for approval before execution.

What makes a software implementation plan grounded instead of guessed?▼

A grounded plan verifies every file path and framework API by reading the actual repository, including AGENTS.md and context maps. It also names contracts at layer boundaries and includes a pre-mortem identifying how the work could silently fail.

Does the plan skill reuse existing plans for similar tasks?▼

Yes, it checks ~/.agents/plans/<repo-name>/ for relevant existing plans and asks whether to reuse or write a new one. Plans prefixed with EXECUTED- are treated as historical records and never block new planning.

How are large features split into phases in an implementation plan?▼

Larger work is split into ordered phases within a single plan file, each with its own tasks, a Task Ownership table, and a verification gate. Execution proceeds one phase at a time, and a failed gate keeps work inside that phase until fixed.

When should code review findings trigger a new plan?▼

When the code-review skill hands over a review file, the plan skill always writes a new plan rather than reusing earlier ones, since earlier plans produced the defects being corrected. The new plan records the review file as its Source.