About DevSpark
Current Release
Current version: v4.2.0
DevSpark is a structured development process for AI coding assistants. It provides 30 stock command prompts plus helper templates and scripts that give any AI agent a repeatable current-truth workflow.
What It Is
- Not a program. Not a subscription. Just markdown files you copy into your project.
- Agent-agnostic. Works with the 18 supported agents listed in the compatibility table, including Claude Code, GitHub Copilot, Cursor, and Gemini CLI.
- Current-truth powered. Projects keep durable code,
.knowledge, and governance aligned while release moves completed planning scaffolding under human-only.archive/YYYY-MM-DD/<topic>/folders. - Multi-app ready. Optionally manage multiple applications in a monorepo with per-app governance, profiles, and scoped commands.
Core Idea
Most AI coding assistants are powerful but unstructured. You get great code generation, but without a consistent process, quality varies and context gets lost between sessions.
DevSpark solves this by giving your AI agent a workflow:
- Specify what you want to build (requirements and user stories)
- Plan how to build it (technical design)
- Break it down into tasks
- Implement with constitution-based guardrails
- Review against your project's own principles
- Release with proper documentation
Design Principles
Markdown-First
Everything is a markdown file. No proprietary formats, no databases, no lock-in.
Temporary specs and plans live only while work is in flight. Durable current
truth lives in code, .knowledge, and governance documents.
Three-Tier Prompt Resolution
Commands resolve through a priority chain:
- Personal overrides —
.knowledge/overrides/{git-user}/commands/— your individual tweaks - Team customizations —
.knowledge/overrides/commands/— shared by the team - Stock defaults —
.devspark/defaults/commands/— the out-of-box prompts
This means teams share a common workflow while individuals can customize any command without affecting others.
Clean Separation
DevSpark keeps its framework files (.devspark/) separate from repository
guides (.knowledge/), current truth (.knowledge/), and ephemeral work
packages (.devspark.work/).
In this source repository, development shims may point directly to the tracked
templates/commands/ files. A consumer installation uses the generated stock
copies under .devspark/defaults/commands/; both layouts resolve commands
through the same personal, team, and stock precedence order.
Who It's For
- Solo developers who want a repeatable process when working with AI assistants
- Teams that need consistent AI-assisted workflows across different agents and editors
- Projects of any size — one-off fix, quick spec, and full spec routes keep the process right-sized
- Monorepo teams — optional multi-app support provides per-application governance, profiles, and scoped commands without requiring any changes to single-app repositories
Origin
DevSpark is an independent project maintained by Mark Hazleton and the open-source community. It provides a structured development process for AI coding assistants, covering the full development lifecycle from requirements through release.