Start from a seed, not a blank prompt.
SeedSpec starts with one SPEC.md. Bundle extra intent and reference material with it, then distribute one package for agents to implement from—instead of copy-pasting prompts.
What’s in a SeedSpec?
SeedSpec starts with just one SPEC.md file and lets you bundle other pieces of intent like skills, evals, behaviors, and more. Nested packages are for optional features or capabilities.
Required
SPEC.mdthe spec
Optional packaging
context-modules/seedspec-engineer/SKILL.mdagent skill
cloudflare-worker/PROFILE.mdimplementation profile
packages/notifications/SPEC.mda nested seed
- Start with the specWrite it like you would for a person.
- Bundle the rest of the intentContext modules cover intent, skill, behavior, evaluation, policy, reference, and implementation profiles.
- Nest what should stay optionalNested packages let you ship extra features or capabilities without rewriting the primary spec.
Why use SeedSpec?
A spec says what to build, but it doesn’t give a typical agent the extra skills or quality checks that make the implementation better.
You bring your own agent, like Claude or ChatGPT. Give it the SeedSpec package, and it has access to the additional intent that guides a better implementation.
See how it feels to use
Two walkthroughs from the spec to the result an agent should produce.
Operations Dashboard Starter Kit
Turn live operations into focused action.A reusable operations console with trustworthy metrics, prioritized records, saved views, governed actions, and safe exports.
- Configured vocabulary
- Metric freshness
- Role-aware actions
- Audit history
See how the package becomes a Returns Operations console through configuration, then how an agent would map it to the real environment.
Customer Onboarding Orchestrator
Make every handoff owned and visible.Turn an approved deal into milestones, owners, customer updates, stalled-work escalation, and auditable completion.
- Three implementation paths
- Milestone ownership
- Escalation
- Evidence
Follow one durable outcome through configured SaaS, an automation platform, or a dedicated coordination service.
Use cases
Share approved product thinking, vendor guidance, or client work without hiding the spec itself.
Internal enterprise libraries
Publish approved product, feature, workflow, and configuration packages with company terminology, policy, guidance, and release expectations.
Explore this use case →Vendor-produced specifications
Package the best way to build with a platform—including its data model, APIs, extension points, examples, and operational guidance.
Explore this use case →Consultancies and agencies
Reuse proven product and domain thinking while adapting every implementation to the customer’s stack, policies, and operating model.
Explore this use case →Read the contract
The spec, schemas, and conformance suite are the contract. The rest of this site is explanation.