hs:plan

Creates verified implementation plans with research, constraint scanning, red-team review, and human approval gates.

Updated Jul 19, 2026
One-click install
npx skills add https://github.com/Dozyboy/VSF --skill hs-plan-dozyboy
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: hs:plan
Source: https://github.com/Dozyboy/VSF/tree/main/Day2_VSF/Demo1/harness/plugins/hs/skills/plan
Command: npx skills add https://github.com/Dozyboy/VSF --skill hs-plan-dozyboy

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Writing code without a verified plan leads to scope creep, unverified assumptions, and rework. This Skill produces a structured, evidence-anchored implementation plan under plans/<timestamp>-<slug>/ that must pass research, constraint-scan, red-team, and validation gates before any code is written. ## Core Features & Use Cases - Gated planning workflow: Runs understand, scope-challenge, research, constraint-scan, plan-writing, red-team, validate, and consistency-sweep steps, with human approval recorded via plan_approval.py before execution. - Mode and flag system: Supports --fast (small tasks), --hard (default, full gates), plus orthogonal --tdd, --deep, --parallel, and --in-place flags that shape phase files and delegation. - Subagent delegation: Delegates research to @researcher, plan-writing to @planner, and adversarial review to @red-teamer, while keeping user-facing interview steps at the main thread. - Use Case: Before refactoring an authentication module, run the plan skill to research the codebase, scan ownership and policy constraints, decompose the work into phase files with a machine-readable plan-graph.yaml sidecar, red-team the design, and get explicit human approval before handing the plan to the cook step. ## Quick Start Ask the AI to create a verified implementation plan for your feature or refactor using the hs:plan skill, optionally with flags like --hard --tdd.

Frequently Asked Questions about hs:plan

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

FAQPage Schema
How do I create an implementation plan before writing code?▼

Run the hs:plan skill with your feature or refactor description. It interviews you for scope and acceptance criteria, researches the codebase, scans ownership and policy constraints, then writes plan.md plus phase files under plans/<timestamp>-<slug>/ for human approval.

What is the difference between --fast and --hard planning modes?▼

--fast is for small, low-risk tasks of 1-2 files and skips research and red-team gates. --hard is the default for real features and runs the full pipeline: constraint scan, red-team review, and validation before approval.

How does the red-team gate review a plan?▼

The red-team gate spawns a @red-teamer subagent that adversarially reviews the plan using 2-4 personas such as Security Adversary and Failure Mode Analyst. Every finding needs file:line evidence, is capped at 15 items, and gets an accept/reject disposition recorded in plan.md.

Can I plan test-driven development phases with the --tdd flag?▼

Yes, --tdd is an orthogonal flag that adds Tests Before, Implement, Tests After, and a Regression Gate section to each phase file. The regression gate records the exact test command so the execution step runs red-to-green in the correct order.

Why does the plan require a plan-graph.yaml sidecar?▼

The plan-graph.yaml sidecar makes the phase dependency graph machine-checkable. The plan_graph.py tool uses it to detect dependency cycles, ordering hazards, and shared-file conflicts between parallel phases, and it hard-fails on any phase missing its post artifact declaration.

When should I not use this planning skill?▼

Skip it for trivial changes where planning overhead exceeds the work, such as a one-line fix with no logic change. The skill itself defaults to --fast for 1-2 file, low-risk tasks, and it never writes code or modifies files outside the plans/ directory.