Skip to content
Nexload SDK

Agent Setup & Quickstart | Nexload SDK

Nexload Agent Skills can be integrated into your workspace either autonomously via our One-Shot Project Adoption Prompt or manually using the official skills CLI.

Choose Your Setup Method

Section titled “Choose Your Setup Method”

Copy the prompt below and send it directly to your AI coding assistant (Cursor, Claude Code, Antigravity, Windsurf, or Codex) inside your target project workspace.

Act as a Principal Systems Architect. Inspect this repository, selectively install matching Nexload Agent Skills ("gecut/nexload-sdk"), integrate our project AGENTS.md, and recommend relevant non-Nexload ecosystem skills via skills.sh.


Execute these 5 phases autonomously:


### PHASE 1: STACK INFERENCE (INSPECT BEFORE ACTING)
Audit repository configuration files (package.json, tsconfig.json, docker-compose.yml, directories, lockfiles):
- Runtime & Engine: Node.js vs Bun | Monorepo vs Standalone.
- Frameworks & UI: Next.js (App/Pages router), React, Tailwind, Payload CMS.
- Reliability & Cloud: Health endpoints, Docker, Kubernetes/Dokploy probes.
- Quality Harness: Test runner (Vitest, Jest), linters, TypeScript strictness.


### PHASE 2: PRUNED SKILL INSTALLATION (COMPLEXITY MUST PAY RENT)
Install ONLY skills with confirmed leverage for this stack using official CLI (`npx skills add gecut/nexload-sdk --skill <name>`):
- Core Reasoning: `nexload-reasoning` (Always install).
  If complex system/refactor: add `nexload-reasoning-discovery`, `nexload-reasoning-investigation`, `nexload-reasoning-ideation`, `nexload-reasoning-design`, `nexload-reasoning-evaluation`, `nexload-reasoning-execution`.
- TypeScript / Library: `nexload-code` (If TS codebase) | `nexload-package` + `nexload-cto-review` (If publishing npm package/monorepo).
- React / Next.js: `nexload-react` + `nexload-design`.
- Reliability / Probes: `healthcheck-core` + runtime adapter (`healthcheck-nextjs-routes`, `healthcheck-node`, or `healthcheck-bun`).
- Payload CMS: `payload-fields-core` + (`payload-editor-core`, `payload-schema-use`, or `payload-operations-core` as needed).


### PHASE 3: NON-DESTRUCTIVE AGENTS.MD BINDING
Create or append to root `AGENTS.md` (never erase existing project conventions):
1. Embed The Inviolable Constitution:
   - Grounding: Runtime > Code > Official Docs > Inferences > Assumptions.
   - Inspect Before Ask: Zero queries for data present in codebase.
   - Problem Space: Feature requests are hypotheses, not constraints.
   - Divergence Firewall: Ideation never self-censors on implementation difficulty.
   - Complexity Rent: Reject speculative scalability; build for Now + 1.
   - Immutability: Never reopen settled ADRs without material contradictory evidence.
   - Verification: "Builds" !== "Works". Claim requires empirical test evidence.
   - Budget: Terminate reasoning when residual uncertainty has zero design impact.
2. Bind explicit triggers for installed skills and define Sparse Handoff (`[NEXLOAD HANDOFF]`).
3. Record exact repo build/lint/test commands.


### PHASE 4: VERIFICATION & AUDIT
Verify installed skill files in `.agents/skills/`. Report:
- Installed skills and precise leverage rationale.
- Intentionally pruned/skipped skills.
- Summary of additions to AGENTS.md.


### PHASE 5: ECOSYSTEM RECONNAISSANCE (SKILLS.SH NON-NEXLOAD SKILLS)
Query `skills.sh` via `npx skills find <stack-keyword>` or web search for non-Nexload community skills matching this stack (e.g., Tailwind, Prisma, Drizzle, Better Auth, Supabase, Vitest, Playwright, Zod, Git):
- Audit top packages matching detected technologies.
- Provide a curated recommendations table: Skill Name, Source (`owner/repo@skill`), Install Command (`npx skills add <source>`), and Target Leverage.
- Ask user confirmation before installing any external skills.

Quick Stack Recipes

Section titled “Quick Stack Recipes”

To avoid installing unneeded skills, use these battle-tested combinations:

Project TypeRecommended Core SkillsWhy This Combination Works
Next.js Full-Stack Appnexload-reasoning
nexload-code
nexload-react
nexload-design
healthcheck-nextjs-routes		Combines structured reasoning with strict React render purity, semantic tokens, and zero-cache health endpoints for Kubernetes or Dokploy.
Node.js / Bun Microservicenexload-reasoning
nexload-code
healthcheck-core
healthcheck-node (or bun)		Ensures boundary narrowing, monitors Linux cgroup resource limits, and prevents container OOM restarts.
TypeScript Library / Monoreponexload-reasoning
nexload-code
nexload-package
nexload-cto-review		Enforces dual CJS/ESM exports, minimal dependencies, bundle discipline, and principal reviewer-only signoff.
Payload CMS Enterprise Appnexload-reasoning
nexload-code
payload-fields-core
payload-editor-core
payload-schema-use		Enforces atomic field validation, deterministic rich-text presets, and single-source-of-truth Zod schema derivation.
Complex Refactor / Migrationnexload-reasoning
nexload-reasoning-discovery
nexload-reasoning-design
nexload-reasoning-evaluation
nexload-reasoning-execution		Full 6+1 cognitive graph pipeline to frame scope, design boundaries, stress-test trade-offs, and verify atomic delivery.

Next Steps

Section titled “Next Steps”
Cognitive Graph (Tier 1) Discover the 6+1 reasoning ecosystem: router, discovery, investigation, and design.
Engineering Standards (Tier 2) Inspect rules for TypeScript boundary narrowing, React purity, and package releases.
Domain Specialists (Tier 3) Browse specialized skills for @nexload-sdk/healthcheck and Payload CMS.
Governance & Protocols Read the 8 inviolable laws of the Constitution and Sparse Pointer Handoff specifications.