Home / baristaze / swe_guidelines · skills/arch-benchmark/SKILL.md · GitHub

arch-benchmark skillA

arch-benchmark is agent-read markdown (skill) from baristaze/swe_guidelines: Run a benchmark scenario (a skill, a command, or a question) in a declared runtime and have frontier models score it against the rubric. Use to measure a change to a skill or the text..

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

# arch-benchmark

Measure a subject against a rubric and report what the judges said.
The harness is `benchmark/run.py` in a checkout of the guideline
repository, and it is the only thing that runs a benchmark. Do not
compose a provider call by hand, and do not invent a flag: every flag
is in `benchmark/README.md`, and `uv run benchmark/run.py --help`
prints them.

## Where it runs

The current working directory must be a checkout of the guideline
repository: `architecture.md`, `.claude-plugin/plugin.json`,
`benchmark/run.py`, and `benchmark/README.md` all present. Check with
`ls` before anything else. Otherwise stop and say so: the benchmark
measures the checkout it runs from, and writes its run folders under
that checkout's `benchmark/runs/`. The copy under
`${CLAUDE_SKILL_DIR}/../..` is the installed plugin, not a checkout;
never run the harness from there.

## Input

`$ARGUMENTS` is one of:

- a scenario name, with or without flags ("explain-tenancy",
  "review-om with all four judges");
- a question about what there is ("which scenarios are there?");
- empty, which means: list the scenarios and the provider
  availability, and stop.

## Procedure
…

Read the whole file at its exact version.

How to install

Latest version
mdr add baristaze/swe_guidelines/arch-benchmark@git:20260922.28cefc1
Exact content
mdr add baristaze/swe_guidelines/arch-benchmark@sha256:2099683d41e27fc4

Pin 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.

Badge

mdr badge

[![mdr](https://markdownregistry.com/badge/art_34wkbao5z7nw4ob7.svg)](https://markdownregistry.com/a/art_34wkbao5z7nw4ob7)

2 badge views in 30 days

Versions

versioncommittedcommitsizeaudit
git:20260922.28cefc1 latest2026-09-22 28cefc1 4,460 BA view · diff
git:20260922.d0b19bf2026-09-22 d0b19bf 4,429 BA view · diff
git:20260922.e4ae97c2026-09-22 e4ae97c 4,173 BA view · diff
git:20260921.419a8e92026-09-21 419a8e9 4,370 BA view · diff
git:20260921.f72eecd2026-09-21 f72eecd 4,383 BA view · diff
git:20260921.93b5e1e2026-09-21 93b5e1e 4,030 BA view · diff
git:20260920.65367e02026-09-20 65367e0 3,921 BA view

Audit of the latest version

A  17 of 17 checks passed. Deterministic, no model, same answer every run.
  • pass: Frontmatter block present
  • pass: Frontmatter declares a name
  • pass: Frontmatter declares a description
  • pass: Size between 200 bytes and 200 KB (4460 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

GitHub

baristaze/swe_guidelines · 86 stars · license MIT · pushed 2026-09-22 · branch main

API

GET https://markdownregistry.com/api/v1/artifacts/art_34wkbao5z7nw4ob7
GET https://markdownregistry.com/api/v1/resolve?ref=baristaze/swe_guidelines/arch-benchmark
GET https://markdownregistry.com/api/v1/blob/2099683d41e27fc4dac601ab507ce57fc46f06d11cc3fe6515060bd9afd327f0

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 baristaze/swe_guidelines

AGENTS.md agents
baristaze/swe_guidelines · AGENTS.md
git:20260922.a3c1e91 · audit A · 86 stars
CLAUDE.md claude
baristaze/swe_guidelines · CLAUDE.md
git:20260915.6033a4d · audit B · 86 stars
arch-benchmark-browser skill
baristaze/swe_guidelines · skills/arch-benchmark-browser/SKILL.md · Run the benchmark prompt in the chatgpt.com, claude.ai, and gemini.google.com products, signed in, and save each answer…
git:20260922.28cefc1 · audit A · 86 stars
arch-deviate skill
baristaze/swe_guidelines · skills/arch-deviate/SKILL.md · Record a deliberate deviation from the guideline as an ADR under docs/adr/: the rule quoted, the decision, the…
git:20260922.81ce1ea · audit A · 86 stars
arch-explain skill
baristaze/swe_guidelines · skills/arch-explain/SKILL.md · Explain how the guideline applies to a question, a file, or a change, citing the sections and lenses that govern it…
git:20260922.e4ae97c · audit A · 86 stars
arch-new-aspect skill
baristaze/swe_guidelines · skills/arch-new-aspect/SKILL.md · Incorporate a new aspect into the guideline and cascade it through the lenses, skills, docs, and README, naming the…
git:20260922.28cefc1 · audit A · 86 stars
arch-review-async skill
baristaze/swe_guidelines · skills/arch-review-async/SKILL.md · Architecture review through the Async lenses: infra, queues, workers, park vs fail. For a change in this area, or as…
git:20260922.28cefc1 · audit A · 86 stars
arch-review-context skill
baristaze/swe_guidelines · skills/arch-review-context/SKILL.md · Architecture review through the Context lenses: stages, scopes, OperatorContext, authorization, tenancy, provenance…
git:20260922.28cefc1 · audit A · 86 stars
arch-review-contracts skill
baristaze/swe_guidelines · skills/arch-review-contracts/SKILL.md · Architecture review through the Contracts lenses: interfaces, injection, wiring. For a change in this area, or as one…
git:20260922.28cefc1 · audit A · 86 stars
arch-review-delivery skill
baristaze/swe_guidelines · skills/arch-review-delivery/SKILL.md · Architecture review through the Delivery lenses: apps, deployment, repo layout, client architecture, logs and…
git:20260922.28cefc1 · audit A · 86 stars
arch-review-full skill
baristaze/swe_guidelines · skills/arch-review-full/SKILL.md · Full architecture review: the eight lens groups of the guideline run in parallel and merge into one report. Use before…
git:20260922.28cefc1 · audit A · 86 stars
arch-review-network skill
baristaze/swe_guidelines · skills/arch-review-network/SKILL.md · Architecture review through the Network lenses: topology, gateway, public types, clients, realtime, push-first. For a…
git:20260922.28cefc1 · audit A · 86 stars

Every file in baristaze/swe_guidelines

Browse by kind, by grade A, or by owner.