Skip to content
/

Panaversity SDD Exercises

panaversity-sdd-exercises · panaversity/claude-code-sdd-exercises · ★ 1 · last commit 2026-02-10

Primitive shape
No installable primitives
00

Summary

Panaversity SDD Exercises — Summary

Panaversity claude-code-sdd-exercises is a pedagogical exercise package, not a framework. It contains 24 exercises across 8 modules plus 3 capstone projects designed to teach Spec-Driven Development concepts using Claude Code or the "Cowork" desktop app. The exercises use real-world planning scenarios (kitchen renovations, charity fundraisers, office moves, weddings) to teach the four-phase SDD workflow: Research → Specification → Refinement → Implementation. No code is written — all exercises are pure planning and problem-solving challenges. The package ships as a flat directory structure with INSTRUCTIONS.md files per exercise, no CLI, no hooks, no commands, and no prompt templates.

Compared to seeds, this package is most similar to agent-os in that it is a methodology document + structured file scaffold rather than a functional framework. Unlike agent-os (which injects working commands into Claude Code), Panaversity SDD Exercises ships only instructional markdown for learners. It is the only pure educational resource in this batch.

01

Overview

Panaversity SDD Exercises — Overview

Origin

Created by Panaversity, tied to "Chapter 5 of the AI Native Development book." Published February 2026. 1 star, 1 contributor.

Philosophy

"Learn to Plan Before You Build"

"SDD is a four-phase workflow that transforms vague ideas into clear, executable plans:

  1. Research — Gather context from multiple angles before deciding anything
  2. Specification — Write a clear document describing what you're building and why
  3. Refinement — Interview stakeholders to find gaps and ambiguities
  4. Implementation — Break the spec into delegatable tasks and execute"

The key pedagogical insight is that SDD skills are domain-agnostic: planning a kitchen renovation or a charity fundraiser teaches the same structural thinking as planning a software feature. By practicing in non-code domains, learners develop the specification habit before adding the cognitive load of programming.

Module Structure

8 modules, 24 exercises, 3 capstone projects, in a recommended 8-week sequence:

  • Week 1 (Why Specs Matter): Module 1 — see spec vs. vibe results side by side
  • Week 2 (Research): Module 2 — multi-perspective research, source reliability
  • Week 3 (Specifications): Module 3 — write first specs from scenarios
  • Week 4 (Constraints): Module 4 — measurable success criteria, adversarial stress-testing
  • Week 5 (Refinement): Module 5 — gap-finding, stakeholder interviews
  • Week 6 (Delegation): Module 6 — dependency mapping, atomic task decomposition
  • Week 7 (Full Cycle): Module 7 — end-to-end SDD on complex scenarios
  • Week 8 (Capstone): Module 8 — apply to personal/professional projects

Target Users

Non-coders and early developers learning to work with AI tools. The README notes: "No coding required — these exercises use real-world planning and problem-solving scenarios."

02

Architecture

Panaversity SDD Exercises — Architecture

Distribution

  • Type: methodology-doc (git clone)
  • License: unknown (not specified)
  • Language: none (pure markdown)

Directory Structure

sdd-exercises/
├── README.md
├── EXERCISE-GUIDE.md           ← Full pedagogical guide
├── module-1/                   ← Spec vs. Vibe: Seeing the Difference
│   ├── exercise-1.1-event-gone-wrong/
│   │   ├── INSTRUCTIONS.md
│   │   └── event-postmortem.md    ← Scenario material
│   ├── exercise-1.2-side-by-side-test/
│   └── exercise-1.3-vibe-to-spec-rescue/
├── module-2/                   ← Research Phase
├── module-3/                   ← Writing First Specs
├── module-4/                   ← Constraints & Success Criteria
├── module-5/                   ← Refinement
├── module-6/                   ← Task Breakdown & Delegation
├── module-7/                   ← Full SDD Cycle
└── module-8/                   ← Capstone Projects
    ├── capstone-A-wedding-planner/
    ├── capstone-B-small-business-launch/
    └── capstone-C-personal-goal-system/

Required Runtime

  • Claude Code (terminal) or Cowork (desktop app)
  • No installation beyond cloning the repo and opening exercises with an AI tool

No Technical Dependencies

No Node.js, Python, or any runtime. Pure markdown files read by Claude or a user.

03

Components

Panaversity SDD Exercises — Components

Exercise Files (24 exercises + 3 capstones)

Each exercise contains:

  • INSTRUCTIONS.md — The exercise brief, task description, reflection questions
  • Scenario material files (postmortems, source documents, voice memos, etc.)

No commands, skills, hooks, agents, scripts, or templates.

Module Breakdown

Module Focus Exercises
1 Spec vs. Vibe (see the difference) 1.1 event disaster, 1.2 side-by-side, 1.3 rescue
2 Research Phase 2.1 multi-angle investigation, 2.2 source reliability, 2.3 research summary
3 Writing First Specs 3.1 home renovation, 3.2 charity fundraiser, 3.3 spec from chaos
4 Constraints & Success Criteria 4.1 missing guardrails, 4.2 measurable vs vague, 4.3 stress-test
5 Refinement: Finding Gaps 5.1 interview challenge, 5.2 ambiguity hunter, 5.3 stakeholder perspectives
6 Task Breakdown & Delegation 6.1 dependency map, 6.2 atomic task writer, 6.3 delegation simulation
7 Full SDD Cycle End-to-End 7.1 community newsletter, 7.2 office move planner, 7.3 product launch
8 Capstone Projects A wedding, B small business, C personal goal system

Pedagogical Techniques

  • Side-by-side comparison: Exercise 1.2 has learners produce two outputs (vague prompt vs spec) and compare
  • Adversarial stress testing: Exercise 4.3 "Constraint Stress Test" — test scenarios designed to break the spec
  • Role simulation: Exercise 5.3 — three stakeholder viewpoints produce three different gap sets
  • Delegation simulation: Exercise 6.3 — assign tasks to three virtual team members
05

Prompts

Panaversity SDD Exercises — Prompt Excerpts

The exercises are prompts given TO Claude by learners, not prompts FROM Claude. The INSTRUCTIONS.md files describe what the learner should ask Claude to do.

Excerpt 1: Exercise 1.1 — Vague vs. Spec Comparison Setup

Source: EXERCISE-GUIDE.md

### Exercise 1.1 — The Event Gone Wrong

**Your Task:**
1. First, read the disaster scenario and ask Claude to "plan a similar event"
   with just a vague prompt
2. Then, write a proper specification for the same event: goals, constraints,
   success criteria, stakeholder needs
3. Ask Claude to plan the event again using your spec
4. Compare both outputs side by side

**What You'll Learn:**
- Vague instructions produce vague results — even from a capable AI
- A specification forces you to think through details BEFORE execution
- The spec doesn't need to be perfect — it just needs to exist

**Reflection Questions:**
1. What details did Claude assume in the vague version that it got wrong?
2. What did your spec force you to decide that you hadn't thought about?
3. How much longer did writing the spec take versus the time it saved?

Prompting technique: Parallel prompt comparison — learners must produce two prompts and compare the outputs, making the specification benefit empirically observable.


Excerpt 2: Module 5 — Ambiguity Hunter

Source: EXERCISE-GUIDE.md

### Exercise 5.2 — Ambiguity Hunter

**The problem:** A spec has been written for a content calendar tool. Your job is to find MORE
ambiguities in the spec than Claude can.

**Your Task:**
1. Read the provided spec
2. Ask Claude to find all ambiguities and gaps
3. Then, independently find more ambiguities Claude missed
4. Document the ones Claude missed

**What You'll Learn:**
- AI models have systematic blind spots in ambiguity detection
- Your domain knowledge finds things AI cannot
- Adversarial review produces better specs than cooperative review

Prompting technique: Adversarial review — learner vs. Claude as competing gap-finders. Teaches that human judgment outperforms AI on domain-specific ambiguity detection.

09

Uniqueness

Panaversity SDD Exercises — Uniqueness

Differs From Seeds

This is the only purely pedagogical resource in this batch. All 11 seeds and all other batch-19 frameworks are functional frameworks; this package is instructional content. It is most similar to agent-os in distribution type (methodology-doc, clone-and-configure) but agent-os installs working Claude Code commands, while Panaversity SDD Exercises ships only INSTRUCTIONS.md files for learners to work through manually.

Positioning

  • Educational supplement to the Panaversity "AI Native Development" book (Chapter 5)
  • Unique in using non-software scenarios (renovations, weddings, fundraisers) to teach software development methodology
  • Targets beginners who have never done formal specification work

Notable Patterns

  1. Domain-neutral SDD: Teaching spec-driven development through life planning scenarios — no code required — is pedagogically unusual
  2. Adversarial comparison (Exercise 5.2): Learner vs. Claude gap-hunting competition
  3. Stress test exercises (Exercise 4.3): Adversarial scenario testing for specifications

Observable Failure Modes

  1. No feedback mechanism: No automated evaluation — learners must self-assess
  2. Dormant maintenance: 1 star, 1 contributor, last commit Feb 2026
  3. Dependency on external tools: Cowork is referenced but Cowork is a separate product not maintained by Panaversity
04

Workflow

Panaversity SDD Exercises — Workflow

SDD Workflow Being Taught

Research → Specification → Refinement → Implementation

Each exercise targets one or more phases, with Module 7 practicing all four end-to-end.

Exercise Workflow

  1. Learner opens an exercise folder
  2. Reads INSTRUCTIONS.md and any scenario material
  3. Applies the target SDD phase with Claude Code or Cowork
  4. Compares output against the exercise's expected learning outcomes
  5. Reflects on the provided reflection questions

No Automation

There are no automated gates, verification scripts, or feedback mechanisms. Evaluation is entirely human or guided by Claude's responses to the learner's work.

Documented in README:

  • Week 1: 1.1, 1.2, 1.3
  • Week 2: 2.1, 2.2, 2.3
  • Week 3: 3.1, 3.2, 3.3
  • Week 4: 4.1, 4.2, 4.3
  • Week 5: 5.1, 5.2, 5.3
  • Week 6: 6.1, 6.2, 6.3
  • Week 7: 7.1, 7.2, 7.3
  • Week 8: Choose one capstone

Approval Gates

None. This is a self-paced learning package.

06

Memory Context

Panaversity SDD Exercises — Memory & Context

No memory or context management. Each exercise is self-contained. There is no state between exercises, no session continuity mechanism, and no persistent store.

Context Model

Learners open an exercise folder, provide the scenario material to Claude manually (or via Cowork), and work within a single session. No cross-session state.

07

Orchestration

Panaversity SDD Exercises — Orchestration

No multi-agent orchestration. Single-agent, human-driven, one exercise at a time.

Execution Mode

One-shot per exercise. Learner runs an exercise, gets output, reflects, moves to next exercise.

Multi-Agent

No — all interaction is between the learner and a single Claude instance.

08

Ui Cli Surface

Panaversity SDD Exercises — UI / CLI Surface

No CLI binary. No web dashboard. No TUI.

Interface

Two options documented:

  1. Claude Code (terminal) — navigate to exercise folder, run claude code, read INSTRUCTIONS.md
  2. Cowork (desktop app) — a separate Anthropic product, not part of this package

Both interfaces are external tools. The package provides only the exercise materials.

Related frameworks

same archetype · same primary tool · same memory type

Context-Engineering Handbook ★ 9.0k

Provides a first-principles, research-grounded vocabulary and learning path for context engineering — the discipline of designing…

walkinglabs/learn-harness-engineering ★ 6.6k

Teach harness engineering from first principles (12 lectures + 6 projects) and provide a scaffolding skill (harness-creator) that…

Awesome Harness Engineering (walkinglabs) ★ 2.7k

Curate the authoritative reference list of articles, benchmarks, and tools for harness engineering — the practice of shaping the…

cline-memory-bank (nickbaumann98) ★ 581

Custom instructions + 6-file hierarchical Markdown memory bank so Cline maintains full project context across sessions, with a…

FPF (First Principles Framework) ★ 372

Provides a formal pattern language for making reasoning explicit, traceable, and publishable in mixed human/AI engineering work —…

nexu-io/harness-engineering-guide ★ 134

Provide a practical, code-first reference guide to harness engineering — from first principles to production patterns —…