How do I install the Seed Unit skill?

Run npx skillmds@latest add inapp-inc/seed-unit in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Seed Unit safe to use?

This skill has not completed SkillMD's automated safety review yet. Capability flags: docs only. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Seed Unit?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Seed Unit free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Seed Unit?

inapp-inc (@inapp-inc) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Seed Unit by inapp-inc — Agent Skill

Name: Seed Unit
Rating: 4.2 (5 reviews)
Author: inapp-inc

Seed Unit

Use to break work into SEED Units (small PR-sized deliverables) with explicit scope, constraints, acceptance criteria, evidence, and rollback. Produces SEED Unit artifacts and ties them to OpenSpec change folders. Invoked by spec-driven-development after architecture.

inapp-inc Updated Sep 22, 2026

File contents

SEED Unit

Convert work into AI-executable SEED Units: small, independently shippable PR-sized slices with explicit outcomes and required evidence.

When: After architecture/design (platform-fit, OpenAPI, container plan decided). Invoked by spec-driven-development.

Core rules

Each SEED Unit MUST have: goal, scope, constraints, acceptance criteria, evidence required, risks & rollback, and a spec link (OpenSpec path).
Keep units small enough for a single PR — except the bootstrap exemption below.
Bootstrap exemption: First greenfield unit (repo/container setup) may exceed single-PR size. Job: invoke scaffold (clone starter after platform-fit), then diverge — not generate scaffolding from nothing.
No hidden work: required work appears in AC and/or tasks.
Under Sprint Mode, compress unit count/granularity; security / performance / observability MUST be filled or N/A — <rationale>.
PR operations are manual.

Invoke engineering skills when constraints apply

Constraint	Invoke
Authn/authz, tenant isolation, validation, secrets, OWASP	security-engineering
Latency/throughput budgets, large data, UI responsiveness	performance-engineering
Runtime behavior needing logs/metrics/traces	observability
Risky rollout / flag / revert	rollback-and-flags

List invoked skills in the SEED Unit block.

Required outputs

SEED Unit block from .cursor/skills/_resources/seed/SEED-UNIT-TEMPLATE.md
OpenSpec change folder: <projectDir>/openspec/changes/<seed-id>/{proposal.md,design.md,tasks.md,specs/**/spec.md}

Engineering checklist (intent-first)

Before implementation: OpenSpec delta exists; AC measurable; constraints explicit; evidence listed; rollback actionable; security/perf/observability filled or N/A.

Use .cursor/skills/_resources/seed/INTENT-REVIEW-CHECKLIST.md.

Dispatch / collection harness

node .cursor/skills/_resources/harness/dispatch/run-dispatch.mjs --tasks "<projectDir>/openspec/changes/<seed-id>/tasks.md"

Prepares scoped prompts and a status table. Does not replace seed-review judgment or launch agents automatically.

---name: seed-unitdescription: Use to break work into SEED Units (small PR-sized deliverables) with explicit scope, constraints, acceptance criteria, evidence, and rollback. Produces SEED Unit artifacts and ties them to OpenSpec change folders. Invoked by spec-driven-development after architecture.---# SEED UnitConvert work into **AI-executable SEED Units**: small, independently shippable PR-sized slices with explicit outcomes and required evidence.**When:** After architecture/design (platform-fit, OpenAPI, container plan decided). Invoked by **spec-driven-development**.## Core rules- Each SEED Unit MUST have: **goal**, **scope**, **constraints**, **acceptance criteria**, **evidence required**, **risks & rollback**, and a **spec link** (OpenSpec path).- Keep units small enough for a single PR — except the **bootstrap exemption** below.- **Bootstrap exemption:** First greenfield unit (repo/container setup) may exceed single-PR size. Job: **invoke scaffold** (clone starter after platform-fit), then diverge — not generate scaffolding from nothing.- No hidden work: required work appears in AC and/or tasks.- Under **Sprint Mode**, compress unit count/granularity; security / performance / observability MUST be filled or `N/A — <rationale>`.- **PR operations are manual.**## Invoke engineering skills when constraints apply| Constraint | Invoke ||------------|--------|| Authn/authz, tenant isolation, validation, secrets, OWASP | **security-engineering** || Latency/throughput budgets, large data, UI responsiveness | **performance-engineering** || Runtime behavior needing logs/metrics/traces | **observability** || Risky rollout / flag / revert | **rollback-and-flags** |List invoked skills in the SEED Unit block.## Required outputs- SEED Unit block from `.cursor/skills/_resources/seed/SEED-UNIT-TEMPLATE.md`- OpenSpec change folder: `<projectDir>/openspec/changes/<seed-id>/{proposal.md,design.md,tasks.md,specs/**/spec.md}`## Engineering checklist (intent-first)Before implementation: OpenSpec delta exists; AC measurable; constraints explicit; evidence listed; rollback actionable; security/perf/observability filled or N/A.Use `.cursor/skills/_resources/seed/INTENT-REVIEW-CHECKLIST.md`.## Dispatch / collection harness```bashnode .cursor/skills/_resources/harness/dispatch/run-dispatch.mjs --tasks "<projectDir>/openspec/changes/<seed-id>/tasks.md"```Prepares scoped prompts and a status table. Does **not** replace seed-review judgment or launch agents automatically.

Frequently asked questions