harness:writing-plans

Generate an end-to-end implementation plan from requirements with acceptance-criteria traceability and scope boundaries.

3|1|Updated Apr 5, 2026
One-click install
npx skills add https://github.com/huangbaixun/harness-engineering --skill harness-writing-plans
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: harness:writing-plans
Source: https://github.com/huangbaixun/harness-engineering/tree/main/skills/writing-plans
Command: npx skills add https://github.com/huangbaixun/harness-engineering --skill harness-writing-plans

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Producing reliable implementations starts with turning a messy spec into a concrete, testable execution plan. This Skill prevents wasted cycles by forcing rigid acceptance-criteria coverage, scoped work boundaries, and bite-sized TDD-ready tasks before any code changes happen.

Core Features & Use Cases

  • Acceptance-criteria coverage gate: Reads features.json and ensures every acceptance_criteria entry maps to at least one task in the plan, preventing orphan requirements.
  • Scope control: Explicitly avoids tasks that would violate features.json out_of_scope constraints, including pushing back when a user asks to expand scope.
  • Implementation-ready decomposition: Produces a file- and step-specific plan with per-task test commands, code blocks, and commit instructions suitable for agentic or inline execution.
  • Architecture decision handling: When a new technical direction emerges during planning, drafts an ADR under docs/decisions/ and references it from the plan.
  • Plan persistence conventions: Saves plans to the harness plugin’s canonical docs/plans location (with support for user override) for consistent retrieval across sessions.

Quick Start

Use the harness:writing-plans skill to create an implementation plan for your next feature by giving it the feature spec and asking it to produce a task-by-task plan that covers every acceptance criterion and avoids out_of_scope work.

Frequently Asked Questions about harness:writing-plans

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

FAQPage Schema
How do I generate an implementation plan from feature specifications?▼

To generate an implementation plan, the system reads a spec context and decomposes multi-step work into file-scoped tasks with specific test commands and commit instructions, ensuring every acceptance criterion is mapped to a task.

How do I enforce acceptance criteria traceability in a TDD execution plan?▼

Acceptance criteria traceability is enforced by reading features.json and verifying that every acceptance_criteria entry maps to at least one test-first execution task in the generated plan, preventing orphan requirements.

What is the best way to manage scope boundaries when planning codebase decomposition?▼

Managing scope boundaries involves explicitly avoiding tasks that violate out_of_scope constraints defined in features.json, including pushing back against user requests that attempt to expand the defined project scope.

When do I need to draft an architecture decision record during implementation planning?▼

An architecture decision record (ADR) is drafted when a new technical direction emerges during the planning phase, saved under docs/decisions/, and referenced directly from the generated plan document.

Does this implementation planning approach support saving plans for retrieval across sessions?▼

Yes, implementation plans are saved to a canonical docs/plans location for consistent retrieval across sessions, with support for user overrides to specify alternative file persistence destinations.