- Home
- Skills
- APIs & Backend
- Bilingual Alexa Custom Skill Scaffold (en-US + es-US)
Bilingual Alexa Custom Skill Scaffold (en-US + es-US)
This kit generates the models from your catalog, checks the package the way certification does, and gives you a handler that behaves the way reviewers test.
$29
Works with the AI Agents you already use
Bilingual Alexa Custom Skill Scaffold (en-US + es-US)
Example session with this skill installed
Build me an Alexa skill in English and Spanish that looks up our 269 public regulations by number (like 300-01) or by topic, and make sure it will pass certification the first time.
- Read your context and instructions
- Compiled the bilingual alexa custom
Summary
Models generated for both locales, package checked, Lambda and template archives built; two things to fix before ask deploy.
What I did
generate-models.mjs --catalog regulations.json --skill-dir .-> en-US and es-US models, 269 slot values each ("three hundred dash zero one" / "trescientos guion cero uno" with digit synonyms), topic search on AMAZON.SearchQuery with carrier phrases.check-skill-package.mjs .-> 2 FAIL, 1 WARN:- FAIL es-US: no example phrase contains the invocation name "buscador de normas" (the phrases were machine-translated from English and kept the English name).
- FAIL en-US: largeIconUri is 500x500; Amazon requires 512x512 PNG.
- WARN: endpoint is the <ALEXA_LAMBDA_ARN> placeholder (fine for a template build).
- Wired
lambda/index.mjswithcreateSkilland your catalog lookup;node --testpasses, including the Spanish refusal for "cómo evadir la inspección" and rejection of a request for another skill id. build-package.mjs --skill-dir .-> lambda.zip and skill-package-template.zip.
Next
Fix the two failures, set ALEXA_SKILL_ID on the Lambda, then build-package.mjs --release --lambda-arn arn:aws:lambda:... and test every example phrase on a device in both languages.
Connects securely to your tools. The creator never sees your data.
About this skill
Alexa skills are rejected for manifests, models and phrasing far more often than for code: a locale without a model, an invocation name containing a launch word, icons at the wrong size, a {query} slot that is the whole utterance, example phrases without the invocation name, Spanish numbers read the English way. This kit generates the models from your catalog, checks the package the way certification does, and gives you a handler that behaves the way reviewers test.
generate-models.mjs: en-US and es-US interaction models from { items: [{ id, title }] }, a reference-number intent whose slot values are the spoken numbers ("three hundred dash zero one", "trescientos guion cero uno") with digit and no-separator synonyms, a topic-search intent on AMAZON.SearchQuery with carrier phrases, required built-in intents, configurable invocation names and samples. number-words.mjs: English and Spanish number words 0..999,999 with the Spanish rules pinned by tests.
check-skill-package.mjs: locale and model agreement, invocation-name rules, three "Alexa, ..." example phrases containing the invocation name, summary and description limits, 108 and 512 PNG icons, explicit privacy booleans, privacy URL when needed, testing instructions, distribution, endpoint shape, sample lowercase and declared slots, AMAZON.SearchQuery misuse, slot-type limits, file sizes, placeholder ARN handling; --release mode.
handler.mjs: createSkill({ skillId, name, lookup }) with skill-id and timestamp verification, launch/help/stop/cancel/repeat/fallback, exact and ranked results with a card of source links, SSML escaping, bilingual strings, and an evidence-bounded refusal filter for security-sensitive questions in both languages. build-package.mjs: Lambda and skill-package zips with a built-in ZIP writer; the release build substitutes the Lambda ARN and refuses a failing package.
Reference: the certification checklist (Vendor ID vs Customer ID, skill id, manifest, model, behavior, evidence-bounded answers, Spanish locale). Templates for the manifest and the Lambda entry point. 6 tests. Node 20+, no packages, no network.
Deployment/IAM is outside scope. Number generation capped at 999,999. HTTPS endpoints require manual signature verification logic. Certification checklist reflects 2026 standards.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 18 days ago
Requires
Permissions
Reads your catalog and skill-package files and writes the generated models and build archives into the skill directory. No network, no environment variables (ALEXA_SKILL_ID is read by your deployed Lambda, not by these scripts).