model-routing skillA
model-routing is agent-read markdown (skill) from nahid-sparktales/agent-dispatcher: Pick a model per job and degrade sensibly when one fails — a quality bar per call site, candidates compared on the same task set, a readable routing rule, and an explicit retry-versus-fallback path with pinned model ids. Use when cost or latency has become a problem, when adding a cheaper or larger model to an existing system, or when a fallback fires silently and quality drops without anyone noticing. Not for prompt authoring, not for retrieval tuning, and not for capacity or infrastructure pla.
Indexed from public GitHub and served as immutable, content-addressed versions. Install it pinned to an exact SHA-256 with the mdr CLI, and every file is verified against the hash recorded here before it reaches your agent. The deterministic audit below grades the latest version, and the same file always earns the same grade.
What the file says
# Model routing Routing is a claim that one job needs less than another. That claim is either measured or it is a guess with a billing line attached. The expensive mistake is not picking the wrong model — it is switching models and having no way to notice what got worse. ## When this fires A system calls more than one model, or is about to, and someone needs to decide which call goes where and what happens when a call fails. It also fires when a fallback exists and nobody can say how often it triggers. It does not fire for a single call site with no cost or latency pressure — there, one capable model and no routing layer is the correct design. ## Procedure 1. **Enumerate the jobs, not the models.** Each call site is a job with its own latency budget, output shape, error tolerance, and whether a human sees the output before it has an effect. Routing decisions belong per job; a single global "which model do we use" question has no answer. 2. **Write the quality bar for each job before looking at any candidate.** What output counts as acceptable, and how that is checked: exact match, schema validity, a rubric, a graded sample, …
Read the whole file at its exact version.
How to install
mdr add nahid-sparktales/agent-dispatcher/model-routing@git:20260919.a0d4f55mdr add nahid-sparktales/agent-dispatcher/model-routing@sha256:66e6859df17db82fPin to a label to follow the author's releases, or to a sha256 to freeze the exact bytes forever. Either way the resolved hash is written to mdr.lock, and mdr install reproduces it on any machine.
[](https://markdownregistry.com/a/art_4xvryyvwnbdluwjo)
1 badge views in 30 days
Versions
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (7633 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
nahid-sparktales/agent-dispatcher · 49 stars · license MIT · pushed 2026-09-23 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_4xvryyvwnbdluwjo GET https://markdownregistry.com/api/v1/resolve?ref=nahid-sparktales/agent-dispatcher/model-routing GET https://markdownregistry.com/api/v1/blob/66e6859df17db82f84ad236e9be15a007084ce20413991e9b54ca79aba16d4b3
Your agent does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.
More from nahid-sparktales/agent-dispatcher
Every file in nahid-sparktales/agent-dispatcher