How long does the integration take?

Change one base URL in an OpenAI-compatible client, or install the Python or TypeScript SDK; the quickstart routes a first request in a few minutes.

Will this lock us into another vendor?

No. The router is Apache-2.0 open source, your provider keys stay yours, and the API is OpenAI-compatible, so leaving is a change of base URL.

What happens when a provider fails?

The router falls back to the next model your policy allows and records the fallback in the trace, so an outage costs a retry, not an incident.

root - Skill - OpenSmartRoute

Name: root
Author: tuandung222

Imported from tuandung222/LuatCheck-D2 (docs/historical/root/AGENTS.md). Install upstream with npx skills add tuandung222/LuatCheck-D2 --skill root. Copyright stays with the author.

LuatCheck integrity rules

Application, demo and experiment entry points must use real configured inference. Never replace an unavailable or malformed model response with a rule-based/mock answer.
A backend failure is an execution error, not a successful answer, a safety refusal, or a zero-cost measurement.
No mocked tests, models, HTTP responses, injected stand-in components, monkeypatching, or canned model actions are permitted. Test real implementations with source-backed data and native inference. Temporary storage is allowed; component substitution is not.
Label lexical scores as lexical proxies. Answer-correctness claims require separate reviewed evidence. Never manufacture scores, token counts, hardware use, human labels, provenance, or member contributions.
Keep unverified historical results out of current report tables. Preserve their historical status.
Missing empirical results in the D2 draft must remain visible red placeholders until supported by replayable run artifacts.
The latest user request authorizes real benchmarking on this Mac and report completion. Preserve failed runs, freeze the selected benchmark before main collection, and never fill missing evidence with fabricated scores or human labels.

Direction: domain-general assistant

The long-term goal is a grounded assistant for any document domain, not only law. The current release covers one domain (the original 2019 Vietnamese Labour Code); keep that scope in claims until another domain is actually evaluated.

New core code must stay domain-agnostic: generic document IDs, tool names and contracts (e.g. src/tools/domain_tools.py, src/contracts/universal_document.py), with domain specifics in data and configuration.
Keep domain-specific parts behind explicit, swappable per-domain modules or config: system prompt, tool names/descriptions (search_law, get_article), ID conventions (Dieu_<n>), guardrail patterns, L1 quantity units, and benchmark tasks/rubrics. Do not add new legal-only assumptions to shared code.
A new domain needs its own source-backed corpus with provenance, its own reviewed tasks and evaluation run before any claim about it. Do not reuse labour-law results as evidence for other domains.

Install root and it becomes one more candidate the router can pick - when it fits.

A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.

{ "ocm": "1", "id": "tuandung222-luatcheck-d2-root", "kind": "skill", "name": "root", "description": "The long-term goal is a grounded assistant for any document domain, not only law. The current release covers one domain (the original 2019 Vietnamese Labour Code); keep that scope in claims until another domain is actually evaluated.", "publisher": "tuandung222", "version": "1.0.0", "capabilities": { "domains": [ "coding", "science" ], "tags": [ "skill-md", "github-agents" ], "languages": [ "en" ] }, "quality_prior": 0.6, "examples": [ "The long-term goal is a grounded assistant for any document domain, not only law. The current release covers one domain (the original 2019 Vietnamese Labour Code); keep that scope in claims until another domain is actually evaluated." ], "primary": false, "metadata": { "source": { "provider": "github-agents", "repository": "https://github.com/tuandung222/LuatCheck-D2", "path": "docs/historical/root/AGENTS.md", "ref": "3c1f29ec2b3ab16a2b099ec2dfe49012136667a1", "url": "https://github.com/tuandung222/LuatCheck-D2/blob/3c1f29ec2b3ab16a2b099ec2dfe49012136667a1/docs/historical/root/AGENTS.md", "key": "tuandung222/LuatCheck-D2/docs/historical/root/AGENTS.md" } }, "instructions": "# LuatCheck integrity rules\n\n- Application, demo and experiment entry points must use real configured inference. Never replace an unavailable or malformed model response with a rule-based/mock answer.\n- A backend failure is an execution error, not a successful answer, a safety refusal, or a zero-cost measurement.\n- No mocked tests, models, HTTP responses, injected stand-in components, monkeypatching, or canned model actions are permitted. Test real implementations with source-backed data and native inference. Temporary storage is allowed; component substitution is not.\n- Label lexical scores a", "cost": { "context_tokens": 561 } }

root

About

LuatCheck integrity rules

Direction: domain-general assistant

Install root and it becomes one more candidate the router can pick - when it fits.

What it costs to adopt - in engineering time.

One base URL

No vendor lock-in

Failover built in

Use it

Manifest

Reviews