agentmods.dev

benchmark-models

benchmark-models is a skill for Claude Code from rjmurillo/ai-agents. It costs 109 tokens per session (2,480 once invoked), scanned A, original, MIT.

A tool for running the same prompt or skill through Claude, GPT, and Gemini and comparing their speed, usage, cost, tool calls, and optional quality scores.

In plain words
What is it for?
Use it to compare model providers, check readiness before a benchmark, and optionally save or judge the results.
Why use it?
It replaces guesswork about which model works best with results from the same task run across multiple models.

Skill for Claude Code

Written for Claude Code: allowed-tools in frontmatter. Also seen: reads .claude/ paths; names the AskUserQuestion tool; positional $N argument.

Part of the project-toolkit plugin — 74 skills shipped together

not rated 49repo +1 changed 6d ago A scan SkillSpector: warn 109 tokens original MIT

Good fit Use it to compare model providers, check readiness before a benchmark, and optionally save or judge the results.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/rjmurillo/ai-agents/benchmark-models
View source ↗ rjmurillo/ai-agents
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add rjmurillo/ai-agents --skill benchmark-models
Clone the repo
git clone --depth 1 https://github.com/rjmurillo/ai-agents

Made for: Claude Code.

Or install project-toolkit, the plugin that ships this one along with the rest of its 74 skills.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for benchmark-models

README.md
[![agentmods](https://agentmods.dev/badge/skills/rjmurillo/ai-agents/benchmark-models/github.svg)](https://agentmods.dev/skills/rjmurillo/ai-agents/benchmark-models)
Your own site
<a href="https://agentmods.dev/skills/rjmurillo/ai-agents/benchmark-models"><img src="https://agentmods.dev/badge/skills/rjmurillo/ai-agents/benchmark-models/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for benchmark-models

Your own site · 80×15
<a href="https://agentmods.dev/skills/rjmurillo/ai-agents/benchmark-models"><img src="https://agentmods.dev/badge/skills/rjmurillo/ai-agents/benchmark-models.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 109 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,480 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector warn 7 Sept 2026
SkillSpector: 1 finding, up to medium

These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →

  • medium Excessive Agency · line 144
    Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
    Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
How audits are shown
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00109 $0.02480
Opus 5.5 $0.00044 $0.00992
Sonnet 5.5 $0.00022 $0.00496
Haiku 4.5 $0.00011 $0.00248

Measured 6d ago against content hash 00414ce626f0, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

benchmark-models scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.

The scan reads SKILL.md. This mod also ships 1 executable file (scripts/model_benchmark.py), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

src/claude/skills/benchmark-models/SKILL.md · 203 lines

How it starts

The opening of the file, as written. The whole thing — 203 lines — stays where its author put it; the contents beside it link to each section on GitHub.

benchmark-models

Run the same prompt across providers and compare them on data, not vibes. The deterministic work lives in scripts/model_benchmark.py (stdlib-only, ported from gstack's bun benchmark; no SDKs, no gstack runtime). This skill is the interactive wrapper: pick a prompt, preview auth with a dry-run, confirm providers, run, interpret, optionally save.

Providers wrap existing CLIs and their own auth (no API keys needed except the optional judge):

  • claude via claude -p --output-format json
  • gpt via codex exec ... -s read-only --json
  • gemini via gemini -p ... --output-format stream-json

Triggers

Invoke this skill when the user says any of:

  • benchmark models
  • compare models
  • model shootout
  • which model is best
  • cross-model comparison

A request for the best model gets per-metric leaders and a rerun plan, not a winner (Step 4).

Do NOT invoke for web page performance (that is a different concern).

Preconditions

  • At least one provider CLI installed and authenticated. The Step 1 dry-run reports exactly which are ready; if none are, the skill stops.
  • The optional quality judge needs ANTHROPIC_API_KEY (adds ~$0.05/run).

Process

Resolve the driver first. In a repository checkout, use BENCH="$(git rev-parse --show-toplevel)/.claude/skills/benchmark-models/scripts/model_benchmark.py". From an installed skill copy, use the local skill path: BENCH="$PWD/scripts/model_benchmark.py" when your shell is in the benchmark-models skill directory.

Step 1: Pick a prompt and preview auth (dry-run)

Decide the prompt with AskUserQuestion:

  • A) Benchmark a skill: pass that skill's SKILL.md path as the prompt. To list candidates, search the installed skills dir, not just . (skills usually live under ~/.claude/skills/ or ~/.copilot/skills/, not the project root).

    Allowed-root constraint: $BENCH reads a prompt file only when it resolves under the repo checkout, the current directory, or a path listed in MODEL_BENCHMARK_SKILL_ROOTS (see resolve_prompt and _allowed_prompt_roots in the benchmark script). A SKILL.md under ~/.claude/skills/ or ~/.copilot/skills/ is outside the repo and cwd, so the run aborts with exit code 2 ("prompt file must be under an allowed root") and benchmarks nothing. Before running, export the skills dir, for example export MODEL_BENCHMARK_SKILL_ROOTS=~/.claude/skills (or ~/.copilot/skills), or copy the file into and run from an allowed root. A prompt already inside the repo or cwd needs no env var.

  • B) Inline prompt: pass --prompt "<text>".

  • C) A prompt file on disk: pass its path (verify it exists first).

Read the full file on GitHub · 203 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 6d ago Changed · +5 lines · +5 tokens per session 00414ce626f0
  2. 14d ago Changed 640d5ef0a2e4
  3. 26d ago Changed d4db0f6592ab
  4. 1mo ago Changed · +5 lines 677cb150219c
  5. 1mo ago Changed · +16 lines 95016358c479
  6. 1mo ago First seen · 177 lines · 104 tokens per session scan A e6dc6102a86b

Subscribe to this mod's changes

benchmark-models is a skill published in the GitHub repository rjmurillo/ai-agents (49 stars, last pushed today), licensed MIT. It adds 109 tokens to every session and 2,480 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.

Analytics and metricsCI and CDClaude CodeCode reviewData and AIagentic-aiai-agentsdeveloper-toolsgithub-copilot
Related

Other skills, from other repositories

data-pipeline-orchestration

Use when adding or changing a scheduled data or ML pipeline — Dagster assets, Prefect flows, Airflow DAGs or a cron'd entrypoint — idempotent partitioned stages, retries, backfills, quality gates, and testing the pipeline without the scheduler.

makifbaysal/tasktrooper · 60 tokens

promptkit

PromptKit composition engine. Use this skill when the user wants to assemble a task-specific prompt, write a requirements doc, investigate a bug, review code, plan an implementation, author agent instructions, create a Copilot prompt file, build an agentic workflow, or perform any engineering task that PromptKit has a…

microsoft/PromptKit · 93 tokens

bootstrap

Alias for the /promptkit skill. Use this when the user wants to assemble a task-specific prompt using PromptKit, or says "bootstrap" to start the PromptKit composition engine.

microsoft/PromptKit · 39 tokens

install-gemini-code-review-action

Installs, configures, or upgrades Dazbo's PR code review & issue triage GitHub Action (gemini-review-action) in a target repository. Use when the user wants to set up, configure, or update automated AI code reviews, issue triaging, or Gemini model workflows for their GitHub repository.

derailed-dash/dazbo-agent-skills · 73 tokens

apex-azure-kusto

ANALYSIS SKILL — Query and analyze data in Azure Data Explorer (Kusto/ADX) using KQL. WHEN: "KQL queries", "Kusto database queries", "Azure Data Explorer", "ADX clusters", "time series data", "IoT telemetry", "anomaly detection". DO NOT USE FOR: App Insights / Log Analytics troubleshooting (apex-azure-diagnostics)…

jonathan-vella/apex-accelerator · 100 tokens

apply-review-feedback

Classify and apply accepted PR/MR review feedback.

juliusz-cwiakalski/agentic-delivery-os · 15 tokens