Borrowing it
Nothing to install: this file belongs to przeprogramowani/10x-bench. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/przeprogramowani/10x-bench/master/.claude/skills/10x-eval-model/SKILL.mdgit clone --depth 1 https://github.com/przeprogramowani/10x-benchWrote 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.
[](https://agentmods.dev/skills/przeprogramowani/10x-bench/10x-eval-model)<a href="https://agentmods.dev/skills/przeprogramowani/10x-bench/10x-eval-model"><img src="https://agentmods.dev/badge/skills/przeprogramowani/10x-bench/10x-eval-model/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.
<a href="https://agentmods.dev/skills/przeprogramowani/10x-bench/10x-eval-model"><img src="https://agentmods.dev/badge/skills/przeprogramowani/10x-bench/10x-eval-model.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00106 | $0.03102 |
| Opus 5.5 | $0.00042 | $0.01241 |
| Sonnet 5.5 | $0.00021 | $0.00620 |
| Haiku 4.5 | $0.00011 | $0.00310 |
Grade E, and why
10x-eval-model scanned grade E with 3 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 27d ago.
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.
Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
**Version:** OpenCode 2.x (`opencode --version`; install/upgrade with `curl -fsSL https://opencode.ai/v2/install | bash`). Runs go through OpenCode's shared background service by default; the session still records the at Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
If runs fail due to credit limits ("requires more credits"), inform the user with a link to https://openrouter.ai/settings/credits and offer to re-run after they top up. Clean attempt directories before re-running (`rm - Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
**Version:** OpenCode 2.x (`opencode --version`; install/upgrade with `curl -fsSL https://opencode.ai/v2/install | bash`). Runs go through OpenCode's shared background service by default; the session still records the at How it starts
The opening of the file, as written. The whole thing — 219 lines — stays where its author put it; the contents beside it link to each section on GitHub.
10x-eval-model
Orchestrates the full setup and execution pipeline for evaluating a new LLM model in the 10xBench benchmark. This includes updating metadata, verifying pricing, preparing attempt directories, and launching evaluation runs via the appropriate harness.
Inputs
The user provides:
- Model ID — the short identifier used in directory names (e.g.,
glm-51,gemini-31-pro) - Display name — human-readable name (e.g.,
GLM-5.1,Gemini 3.1 Pro) - Harness — the coding agent environment used to run the model (e.g.,
opencode,cursor,claude-code,claude-desktop,codex-desktop) - Supersedes (optional) — the model ID this new model replaces (e.g.,
glm-5is superseded byglm-51)
If the user provides a natural model name like "GLM-5.1 via opencode", derive the model ID yourself (e.g., glm-51) and confirm with the user before proceeding.
Pipeline Steps
Step 1: Check if model already exists in metadata
Read eval-attempts/metadata.ts and check whether the model ID is already present. If it exists but is commented out, offer to uncomment it instead of adding new entries. If it already exists and is active, inform the user and skip to Step 4.
Step 2: Look up pricing
Search the web for current API pricing for the model (per 1M tokens, input and output, USD, no cache). Present findings to the user for confirmation before writing to metadata. Check multiple sources (official docs, OpenRouter, Artificial Analysis) to cross-reference.
Step 3: Update metadata.ts
Add the model to all relevant sections in eval-attempts/metadata.ts:
ModelIdtype union — add| "model-id"entryAGENT_NAMES— add"model-id": "Display Name"entryAGENT_ENV— add"model-id": AGENT_ENVIRONMENT.<harness>entry (see harness mapping below)MODEL_PRICING— add"model-id": {input: X, output: Y}with confirmed pricingSUPERSEDED_MODELS(if applicable) — add"old-model-id": "new-model-id"entry
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.
- 27d ago Changed · +196 lines · +51 tokens per session scan A → E de801554c4bd
- 1mo ago Changed · -166 lines · -51 tokens per session scan C → A 890dc31640b3
- 2mo ago First seen · 189 lines · 106 tokens per session scan C 5496a43ed928
10x-eval-model is a skill published in the GitHub repository przeprogramowani/10x-bench (14 stars, last pushed 20d ago), licensed MIT. It adds 106 tokens to every session and 3,102 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it E with 3 findings (downloads and executes remote code, recursive force delete, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
spark-environment-setup
Set up a working ML training/inference environment on NVIDIA DGX Spark (GB10, aarch64, CUDA 13). Use when installing PyTorch/Unsloth/TRL/vLLM on DGX Spark, hitting libcudart or wheel-ABI errors on aarch64, or choosing between NGC containers and bare pip installs.
spark-training-gotchas
Preflight and diagnose the ten known failure modes for ML training on NVIDIA DGX Spark. Use when a training run on DGX Spark fails to start, OOMs below the 128GB limit, slows down mid-run, or before any multi-hour training job on GB10.
9router-stt
Speech-to-text via 9Router /v1/audio/transcriptions using OpenAI Whisper / Groq / Gemini / Deepgram / AssemblyAI / NVIDIA / HuggingFace models. Use when the user wants to transcribe audio, convert speech to text, or get subtitles from audio files.
9router
Entry point for 9Router — local/remote AI gateway with OpenAI-compatible REST for chat, image, TTS, embeddings, web search, web fetch. Use when the user mentions 9Router, NINEROUTERURL, or wants AI without writing provider boilerplate. This skill covers setup + indexes capability skills; fetch the relevant capability…
9router-embeddings
Generate vector embeddings via 9Router /v1/embeddings using OpenAI / Gemini / Mistral / Voyage / Nvidia / GitHub embedding models for RAG, semantic search, similarity. Use when the user wants embeddings, vectors, RAG, semantic search, or to embed text.
dashscope
DashScope (Alibaba Cloud Bailian / 阿里云百炼) integration — image generation (qwen-image-2.0-pro), text-to-speech (qwen3-tts-flash), and ASR with word-level timestamps (qwen3-asr-flash-filetrans). Use when generating images via Qwen-Image, narrating via Qwen-TTS, or transcribing with word-level timestamps via Qwen-ASR.