DevSpark

A structured development process for AI coding assistants.
Just markdown files — no install required.

Live Site: https://dev.makeboldspark.com Current version: v4.2.0

Not a program. Not a subscription. Copy 30 stock command prompts plus the helper templates and scripts into your project and your AI coding assistant gets a repeatable current-truth workflow. Works with the 18 supported agents listed below.

---

What's In This Repo

devspark/
├── agents-registry.json  ← Canonical metadata for supported agent integrations
├── templates/commands/   ← 30 stock command prompt files (THE PRODUCT)
├── scripts/              ← Context-gathering scripts (PowerShell + Bash)
├── .knowledge/           ← Current truth: entities, governance, ontology reports
└── .devspark.work/       ← Temporary lifecycle work products

DevSpark Vocabulary

• Prompt: A workflow command surface, usually a slash-command file, that orchestrates DevSpark lifecycle behavior such as specification, planning, implementation, review, or release.
• Agent: An AI runtime or client integration such as Codex, Claude, Copilot, Cursor, or Gemini. Agent metadata lives in agents-registry.json and describes supported integrations, not team responsibilities.
• Skill: A reusable portable capability package that a prompt can delegate to when specialized task knowledge is useful.
• Participant: A human or AI-filled team member carrying responsibility for work, review, approval, critique, or decision capture in a workflow.
• Role: A responsibility label assigned to a participant, such as owner, planner, implementer, reviewer, critic, or scribe.

Participant metadata is optional advisory context in artifacts. It does not change prompt resolution, script resolution, command behavior, or the existing customization process.

Agent Skills

DevSpark treats skills as portable capability packages within a governed lifecycle orchestration system. Commands invoke skills; DevSpark governs the lifecycle around them.

The dual-surface model:

command -> adapter -> skill -> context scripts -> agent reasoning -> artifact

• Slash-commands (templates/commands/) own DevSpark-specific lifecycle behavior: route classification, branch creation, multi-app scoping, artifact placement, and gate enforcement.
• Agent Skills (templates/skills/) own portable capability instructions that run in any skills-compatible client without DevSpark installed.

The adapter contract (templates/skills/ADAPTER-contract.md) defines how a command invokes a skill. The skill validation contract (templates/skills/SKILL-validation-contract.md) defines the rules every SKILL.md must satisfy. See templates/skills/references/devspark-skills-guide.md for the contributor walkthrough for adding new skills.

---

Get Started

DevSpark is installed, upgraded, and repaired only through quickstart prompts. Point your AI agent at the quickstart prompt for your platform:

• GitHub Copilot
• Claude Code
• Cursor
• Codex
• Any other agent

For ongoing updates or repairs, run the same quickstart prompt again in the target repository. The quickstart compares the installed version, refreshes framework-owned files, repairs missing stock assets, initializes missing .knowledge/entities/ and .knowledge/ontology/ scaffolding, and preserves authored repository-owned .knowledge/ content.

For the governing principles, see the DevSpark Philosophy. For a full walkthrough see the Implementation Lifecycle Guide.

Canonical delivery loop: specify → clarify when needed → plan → tasks → required gates → implement → focused verify when needed → commit/push → create-pr → pr-review ↔ address-pr-review → merge. Run release separately when the team is ready to validate and archive completed packages.

---

Slash Commands

Core Workflow

Command	Purpose
/devspark.constitution	Establish project principles and guidelines
/devspark.specify	Define what you want to build (requirements & user stories)
/devspark.plan	Create a technical implementation plan
/devspark.tasks	Break the plan into actionable task lists
/devspark.implement	Execute tasks and build the feature
/devspark.verify	Verify behavioral proof and reject metric-only fixes
/devspark.create-pr	Draft or update a spec-aware pull request
/devspark.update-pr	Refresh an existing pull request description from the current branch delta
/devspark.next	Continue from detected repository, artifact, gate, and PR state

Review, Release, and Repository Utilities

Command	Purpose
/devspark.pr-review	Constitution-based PR review
/devspark.address-pr-review	Apply PR review fixes with mandatory commit isolation gates
/devspark.site-audit	Comprehensive codebase audit
/devspark.explain	Explain existing functionality and verify matching current truth
/devspark.quickfix	Lightweight workflow for bug fixes
/devspark.fix-score	Diagnose and remediate repository score blockers without weakening scoring rules
/devspark.critic	Adversarial risk analysis
/devspark.release	Validate current truth and archive completed work at release time
/devspark.evolve-constitution	Propose constitution amendments
/devspark.repo-story	Generate narrative from commit history
/devspark.commit-audit	Analyze commit history for workflow, hygiene, and delivery signals
/devspark.taskstoissues	Convert tasks.md into dependency-ordered GitHub issues

Quality & Personalization

Command	Purpose
/devspark.clarify	Ask structured questions to de-risk ambiguity
/devspark.analyze	Cross-artifact consistency check
/devspark.checklist	Generate quality validation checklists
/devspark.personalize	Create per-user command overrides
/devspark.discover-constitution	Generate a constitution from existing code
/devspark.discover-knowledge	Build source-grounded .knowledge/entities and ontology

Multi-App (Optional)

Command	Purpose
/devspark.add-application	Register a new application in the multi-app registry
/devspark.list-applications	Display all registered applications
/devspark.validate-registry	Validate registry schema, references, and consistency

See .knowledge/entities/product-documentation/site/index.md for full command details.

---

Multi-App Monorepo Support (Optional)

Single-app repositories need nothing here. Multi-app is entirely optional — if your repo has one application, skip this section entirely. Everything works out of the box.

For repositories containing multiple applications with different platforms, runtimes, or governance rules, DevSpark offers opt-in multi-app support:

When to Use Multi-App

Scenario	Recommendation
Single application or library	Skip multi-app — standard DevSpark is all you need
Monorepo with shared conventions	Skip multi-app — one constitution covers everything
Monorepo with different platforms (e.g., .NET API + React UI)	Consider multi-app — each app can have tailored rules
Monorepo with different governance (e.g., PCI service + internal tool)	Use multi-app — app-specific constitutions and profiles

How It Works

1. Create a registry at .knowledge/entities/application-registry/registry.json — or run /devspark.add-application to create one interactively
2. Assign profiles — reusable rule bundles (e.g., api-profile, web-profile) that apps inherit
3. Scope commands — use --app <id> to target a specific application, or --repo-scope for repo-wide operations
4. App-local overrides — each app can have its own .knowledge/ directory and optional app.json manifest

Multi-App Commands

Command	Purpose
/devspark.add-application	Register a new application in the registry
/devspark.list-applications	Display all registered applications and profiles
/devspark.validate-registry	Validate registry schema, references, and consistency

For the full specification, see the Monorepo Guide.

---

Customization Without Risk

DevSpark cleanly separates your work from its installation:

.devspark/                 ← Installation (removable, upgrade-safe)
├── defaults/commands/     ← Stock prompts
├── scripts/              ← Stock helper scripts
├── templates/            ← Spec/plan templates
└── VERSION               ← Installed version stamp

.knowledge/                ← Repository-owned current truth
├── governance/constitution.md
├── governance/decisions/
├── entities/
├── overrides/commands/             ← Team command overrides
├── overrides/scripts/              ← Team script overrides (optional)
└── overrides/{git-user}/commands/  ← Personal overrides

.devspark.work/            ← Ephemeral in-flight work packages

Multi-app layout (optional): Each registered application may use {app-path}/.knowledge/ for app-local entities, decisions, ontology output, and overrides, plus {app-path}/.devspark.work/ for temporary work packages.

3-tier prompt resolution (first match wins):

1. .knowledge/overrides/{git-user}/commands/ — Personal tweaks
2. .knowledge/overrides/commands/ — Team customizations
3. .devspark/defaults/commands/ — Stock prompts

2-tier script resolution (first match wins):

1. .knowledge/overrides/scripts/ — Team script overrides (e.g., Azure DevOps adapter)
2. .devspark/scripts/ — Stock scripts

There is no third ownership tier. If an organization wants a shared baseline in .knowledge/, it manages that through its own repo practices; DevSpark still only writes to .devspark/.

Participant metadata uses these existing repository-owned artifacts when present. It does not change how prompts or scripts are found; customization layers and precedence are unchanged.

Clean removal: remove .devspark/ and generated agent shims. Leave .knowledge/ intact unless the user explicitly asks to remove repository-owned current truth.

---

Supported AI Agents

DevSpark is agent-agnostic. Every agent below gets thin shims that resolve personal overrides, team overrides, and stock prompts through the same command contract.

Agent		Agent		Agent
Claude Code	✅	Cursor	✅	Gemini CLI	✅
GitHub Copilot	✅	Codex CLI	✅	Windsurf	✅
Amp	✅	Roo Code	✅	Kilo Code	✅
Auggie CLI	✅	opencode	✅	Qwen Code	✅
SHAI	✅	Amazon Q	⚠️	IBM Bob	✅
CodeBuddy	✅	Qoder CLI	✅	Antigravity	✅

---

Learn More

Topic	Link
Implementation lifecycle	implementation-lifecycle.md
Quickstart	quickstart.md
Constitution guide	constitution-guide.md
Installation	installation.md
Upgrading	upgrade.md
PR review guide	pr-review-usage.md
Site audit guide	site-audit-usage.md
Critic guide	critic-usage.md
Repo story	repo-story-usage.md

---

Prerequisites

• Any OS (Linux / macOS / Windows)
• A supported AI coding agent
• Git (recommended)

Contributing

See CONTRIBUTING.md and CODE_OF_CONDUCT.md.

Acknowledgements

DevSpark is maintained by Mark Hazleton and the open-source community.

Built by Mark Hazleton — Mark Hazleton, Solutions Architect DevSpark is part of the Make Bold Spark portfolio of technical demonstrations.

A Make Bold Solutions project, part of the Make Bold Spark family.

License

MIT — see LICENSE.