A research-first scouting agent skill. Maximize long-term business value before you add code or dependencies.
AI coding agents start by writing code. This skill starts with a different rule: evaluate long-term business value before you create new code. Search existing solutions. Compare them on reliability, strategic value, adaptability, total cost of ownership, and speed to value. Then choose the tier that maximizes value for your situation.
The skill now uses a required output schema. Each recommendation includes context, tier search, evidence matrix, recommendation, reasons, confidence, re-evaluation trigger, and missing information.
Unlike cost-only build-vs-buy analyzers, this skill evaluates five components of business value. A commodity (like auth or PDF generation) wins on TCO and speed — USE or BUY the best seller. A core competency (like a matching engine or recommendation system) wins on strategic value and adaptability — BUILD or FORK. The framework has no preferred answer. Your situation determines the right tier.
Start at the top. Move down only when a step finds no viable candidate.
REUSE — a solution exists in the repository or the organization.
USE — a maintained package solves it, installed or added normally.
FORK — no maintained package fits. Copy, vendor, or patch open source.
BUY — a SaaS product solves it.
INTEGRATE — an API solves it.
BUILD — viable when no earlier tier fits. Cite the search.
The order is a search order. It is not a preference order. Your situation determines which tier maximizes value.
Skip the ladder for small local changes. Do not run it for bug fixes, single-line fixes, existing imports, routine refactors, docs, style, or lint.
npx skills add fahrellgiovanny/nothing-new-under-the-sungh skill install fahrellgiovanny/nothing-new-under-the-sun nothing-new-under-the-sun --agent <agent-id> --scope userCopy SKILL.md to the skills directory of your agent.
| Agent | Path |
|---|---|
| OpenCode | .opencode/skills/nothing-new-under-the-sun/SKILL.md |
| Claude Code | .claude/skills/nothing-new-under-the-sun/SKILL.md |
| Codex CLI | .agents/skills/nothing-new-under-the-sun/SKILL.md |
| Cursor | .cursor/skills/nothing-new-under-the-sun/SKILL.md |
| Gemini CLI | .gemini/skills/nothing-new-under-the-sun/SKILL.md |
| GitHub Copilot | .github/skills/nothing-new-under-the-sun/SKILL.md |
| Windsurf | .windsurf/skills/nothing-new-under-the-sun/SKILL.md |
| Cline | .cline/skills/nothing-new-under-the-sun/SKILL.md |
| Roo Code | .roo/skills/nothing-new-under-the-sun/SKILL.md |
| Amp | .agents/skills/nothing-new-under-the-sun/SKILL.md |
| OpenHands | /add-skill https://github.com/fahrellgiovanny/nothing-new-under-the-sun/tree/main/skills |
| Hermes Agent | ~/.hermes/skills/nothing-new-under-the-sun/SKILL.md |
| OpenClaw | ~/.openclaw/skills/nothing-new-under-the-sun/SKILL.md |
| Kiro CLI | ~/.kiro/skills/nothing-new-under-the-sun/SKILL.md |
| Factory Droid | .factory/skills/nothing-new-under-the-sun/SKILL.md |
Agents without native skill directories (Aider, for example) can read the
prompt with /read <path/to/SKILL.md>. The Markdown prompt works with
any LLM.
Copy agents/nothing-new-under-the-sun.md to the agent directory of your
platform. The agent prompt asks for read-only research, and Claude Code and
OpenCode honor the permission: block. On other hosts, configure host-level
sandboxing, hooks, or permissions to deny writes and package installs.
| Host | Skill support | Agent frontmatter | Permission model | User action |
|---|---|---|---|---|
| Claude Code | Native or manual | Yes | Honors permission: |
Configure sandboxing for sensitive repos. |
| OpenCode | Native or manual | Yes | Honors permission:. Last matching bash rule wins. |
Keep "*": ask before specific deny rules. |
| Other listed hosts | Varies | Varies | Do not assume permission: works. |
Configure host sandboxing, hooks, or permissions. |
The skill directs the agent through six search stages:
- Search the repository. The solution may already exist in the codebase.
- Search maintained packages. Check the project manifest, framework features, built-in platform features, and public registries such as npm, PyPI, crates.io, and similar registries.
- Search open-source projects to copy or vendor. Use this only when no maintained package fits. Examine stars, contributors, release rate, issue response time, and license.
- Search SaaS products. Some problems fit a seller product.
- Search APIs. Some problems need integration, not code.
- Check the license. Verify compatibility with the project at every step.
The agent fills an evidence matrix for every candidate. The row below is illustrative. Replace it with real candidates for each request.
| Option | Reliability | Strategic value | Adaptability | TCO | Speed to value | Citation |
|---|---|---|---|---|---|---|
| express-rate-limit | Maintained OSS middleware | none — commodity | Express-only, configurable | LOW (MIT, free) | S (< 1 day) | web_fetch: https://github.com/express-rate-limit/express-rate-limit on <today> |
The agent then returns one recommendation with this required schema:
- Context.
- Tier search.
- Evidence matrix.
- Recommendation.
- Why this wins.
- Confidence.
- Re-evaluation trigger.
- Missing information.
# Request
Should we build our own auth, or use Auth0?
# Recommendation (compound)
USE next-auth for the commodity slice. Small BUILD layer over next-auth
callbacks for enterprise SSO, only when the first enterprise customer
asks for it.
# Why this wins
- Reliability: next-auth proven at scale (5M weekly, no critical CVE).
- Strategic value: identity is a commodity. Engineering effort goes to
product work.
- Adaptability: adapters cover OAuth, passkeys, SAML.
- TCO: near zero license; no per-MAU bill.
- Speed to value: one week to production.
# Request
Should a ride-hailing company build or buy its driver-rider matching
algorithm?
# Recommendation (compound)
BUILD the matching engine. INTEGRATE routing and mapping APIs.
# Why this wins
- Reliability: proven only through iteration; every competitor faced the
same risk.
- Strategic value: matching IS the product. Only BUILD supplies this.
- Adaptability: BUILD gives full control; strategy changes weekly.
- TCO: high build cost but zero seller cost; no lock-in.
- Speed to value: 6-12 months to MVP; compounding gains over 10 years.
This skill follows the Agent Skills specification. It works with any compatible agent.
Compatible agents:
- OpenCode
- Claude Code (Anthropic)
- Codex CLI (OpenAI)
- Cursor
- GitHub Copilot
- Gemini CLI (Google)
- Windsurf
- Cline
- OpenHands
- Hermes Agent (Nous Research)
- OpenClaw
- Roo Code
- Amp
- Kiro CLI
- Factory Droid
For agents without native skill support, paste the prompt or use the
/read command.
| Skill | Agent | |
|---|---|---|
| Loads when | Before a build proposal, automatically | When invoked or dispatched |
| Writes code | Yes, after a justified BUILD recommendation | No |
| Use for | Daily development sessions | Auditable decisions |
| Permissions | Set in the host platform | Set in the agent definition |
Use the skill for daily work. Dispatch the scouting agent when you need an auditable research prompt and host-level read-only controls.
AI coding agents write code first. They produce 50 lines for a problem that one import solves. They build an auth system when three mature libraries exist. They build a service when an API already does the job.
This skill inverts that default. It tells the agent: evaluate long-term business value before you create new code.
The philosophy: there is nothing new under the sun. Your problem is not unique. Existing work covers most of it. Find that work and compare its value before you build.
nothing-new-under-the-sun/
├── README.md
├── LICENSE
├── CHANGELOG.md
├── .pre-commit-hooks.yaml
├── skills/
│ └── nothing-new-under-the-sun/
│ └── SKILL.md
├── agents/
│ └── nothing-new-under-the-sun.md
├── schemas/
│ ├── recommendation.schema.json
│ └── examples/
│ └── valid-recommendation.json
├── templates/
│ └── recommendation.md
└── tests/
└── scenarios.md
This skill follows a long tradition of research-first and build-versus-buy analysis. Direct influences:
- The Agent Skills specification and ecosystem.
- Build-versus-buy analysis in enterprise architecture.
- Prior-art search in patent and open-source contexts.
- Dependency-check-first workflows in modern package managers.
- Wardley Mapping and DDD Core/Supporting/Generic classification.
Related skills use different approaches. This skill differs from all of them. It requires a six-tier recommendation, a five-component evidence matrix, and strict citation format.
Enterprise-grade weighted-scoring tools (Agensi Build-vs-Buy Analyzer) and strategic modeling frameworks (Wardley Mapping) exist. This skill targets a different moment: the instant before an AI agent or engineer writes new code. It runs in one turn and produces a defensible recommendation.
Report issues and request features at github.com/fahrellgiovanny/nothing-new-under-the-sun/issues.
- The skill and agent prompts are Markdown only. They contain no scripts,
executables, or runtime code. The
schemas/andtemplates/files are data and configuration only. - WARNING: Agents run with full platform permissions. Configure each host to deny write, edit, and package install access during research.
- Do not run candidate install commands during research.
- Request approval before you install, copy, or import a candidate.
- Examine the security audit results on skills.sh before you install.
MIT — see LICENSE.
"The thing that hath been, it is that which shall be; and that which is done is that which shall be done: and there is no new thing under the sun." — Ecclesiastes 1:9 (KJV)