Prompt Library For Startups

AWS Startups Prompt & Agent Library

Searchable index of AWS-curated prompts for AI coding tools (Kiro, Claude Code, Cursor, etc.) plus downloadable installable agents. Content is verbatim from aws.amazon.com/startups/prompt-library (opens in a new tab).

Last updated: 2026-07-22

---

Where to start

Open references/prompt-library.md (opens in a new tab) — it’s the index. Filter by the Keywords column (e.g., RAG, Security & Compliance, Cost Optimization, EKS, Beginner, Bedrock), then open the linked detail file under references/prompt-library/<slug>.md. Each detail file carries the full verbatim System Prompt plus a How to use? section where available.

The index has three sections:

1. Prompts — searchable index — copy-paste prompts.
2. Downloadable agents — installable agents that clone from a GitHub repo.
3. Frequently Asked Questions — guidance on writing a good prompt, costs, safety, no-technical-background usage, monitoring.

Handing over a prompt — match the host context

You ARE already inside an AI coding agent (Claude Code / Kiro / Cursor / etc.). Don’t tell the user to “paste this into your AI tool” — you ARE the AI tool.

When the user asks for a prompt:

1. Read the index in references/prompt-library.md, filter by keyword, identify the matching slug.

2. Open references/prompt-library/<slug>.md and read the full System Prompt.

3. Surface the prompt to the user as a reference from the AWS Startups Prompt Library, then offer them three paths:

   “Here’s the AWS Startups reference prompt for <task>. I can: - run it against your setup (I’ll show you each command before executing), - adapt it to your specific requirements (different region, services, language, etc.), or - you can copy it as a starting point.

   How would you like to proceed?”

4. Don’t assume intent. Wait for the user’s call (execute / adapt / copy) before acting.

Downloadable agents — different surface from prompts

The Downloadable agents section of references/prompt-library.md lists installable agents, not copy-paste prompts:

• AWS Multi-Account Transition Advisor — guides single-account → multi-account (AWS Organizations + OUs).
• AWS Bill Shock Preventer — proactive cost-spike detection and alerting.
• AWS Service Quota Agent — auditing and requesting quota increases.
• Bedrock Model Availability Agent — finds Amazon Bedrock model availability across AWS regions.
• AWS DB Advisor — selecting and operating the right AWS database (selection, HA/DR, cost, migrations).

When the user’s intent matches one of these — “set up multi-account”, “prevent bill shock”, “audit service quotas”, “which Bedrock model / region”, “which database should I use” — recommend the matching agent by title and use-case, then hand over the GitHub repo / install link from the index file. Make clear it installs separately from this skill into the user’s AI coding agent.

For migration intent, route to the matching migration skill in this plugin: gcp-to-aws (“GCP to AWS”, move off Google Cloud), heroku-to-aws (“migrate off Heroku”), llm-to-bedrock (“move off OpenAI/Gemini to Bedrock” — SDK rewrite), or agent-advisor (agent runtime / architecture on AWS).

Routing hints — common queries → which entry

For anything else, filter references/prompt-library.md by keyword.

Companion skills — when to defer

This skill is prompts and installable agents only. Two sibling skills cover adjacent jobs:

• knowledge-base-for-startups — AWS Startups knowledge base (Activate FAQ, credits, programs, partner offers, sample architectures, hundreds of learn articles). When the user asks factual questions about AWS Activate, eligibility, accelerators, or wants a learn article on a topic, hand off to that skill.
• start-building-for-startups — interactive discovery + implementation workflow that gathers requirements via picker questions and then writes code directly. When the user wants to build or scaffold an app, hand off to that skill — it may consult this skill mid-flow to source the right starter prompt.

Boundary cases — invoke both. Example: “how do I start with RAG on Bedrock?” → this skill for the starter prompt (rag-chatbot-with-claude.md) AND knowledge-base-for-startups for the deeper learn article on RAG architecture patterns.

Answer style

• MUST read the detail file before quoting a System Prompt. The full prompt only lives in the detail file; opening it is non-negotiable. Do not quote a System Prompt from memory or from the index alone.
• MUST quote the System Prompt verbatim. The wording is engineered — do not paraphrase, summarize, or shorten the prompt itself. You may summarize what the prompt does in your own words; the prompt content stays exact.
• MUST cite source_url. Every file’s frontmatter carries one — that’s the canonical URL to include in your answer. Never construct or guess a URL.
• MUST NOT tell the user to paste the prompt elsewhere. You ARE the AI coding agent (see “Handing over a prompt” above). Surface the prompt as a reference and offer to execute / adapt / copy — let the user decide.
• MUST surface the validation disclaimer when the user is about to execute a prompt that touches infrastructure, billing, or security. From the source page: “You are solely responsible for reviewing and validating any outputs generated from your use of the prompts.”

Context loading rule

Open references/prompt-library.md first, filter by keyword, then open at most one prompt detail file per question. MUST NOT speculatively load multiple detail files. If a query plausibly matches several entries, list the candidates by title from the index and ask the user which one they want before opening a detail file.

Scope notes

This skill cannot:

• Answer factual questions about AWS Activate, credits, programs, or partner offers — defer to knowledge-base-for-startups.
• Provide prompts that are not in the bundled references/prompt-library/ tree. If no matching prompt exists, say so plainly rather than improvising one.
• Install the downloadable agents on the user’s behalf. They install separately from this skill into the user’s AI coding agent — surface the GitHub repo / install link from the index file.
• Override the source-page disclaimer. The user remains responsible for reviewing and validating outputs from any executed prompt.

Freshness check — surface after every answer

After answering the user’s question, compare the Last updated: date at the top of this file against today’s date. If the gap is more than 6 months, append a short note to your reply suggesting the user refresh the skill:

“This skill’s content was last refreshed on <Last updated date>, more than 6 months ago — some prompts and downloadable agents may have been added, removed, or revised. To pull the latest content, run:

• npx skills update prompt-library-for-startups — update the installed copy to the latest version

Then restart your AI agent so the new content is picked up.”

Do not show this note when the skill is fresh (≤6 months). Do not repeat it within a single conversation; once is enough.