# Skippy Spec Bench

> Use this skill when testing or benchmarking target/draft GGUF pairs for speculative decoding compatibility, tokenizer agreement, draft acceptance rate, or staged verification behavior.

- Skill: `mesh-llm/skippy-spec-bench` (Agent Skill)
- Install (CLI): `npx skillmds@latest add mesh-llm/skippy-spec-bench`
- Raw SKILL.md: https://api.skillmd.com/api/skills/mesh-llm/skippy-spec-bench/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: mesh-llm (https://skillmd.com/u/mesh-llm)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/mesh-llm/skippy-spec-bench

---


# skippy-spec-bench

Use this skill for target/draft speculative compatibility work.

## What It Checks

- Target and draft tokenization agreement.
- Baseline target decode versus draft-verified decode.
- Draft acceptance/rejection behavior.
- Batched verification and checkpoint/restore behavior.
- Recurrent-state implications for rollback.

## Repo Notes

The old source repo used a standalone `llama-spec-bench` crate. It may not be
present in this mesh checkout yet, so verify available packages before running
commands:

```bash
cargo metadata --no-deps --format-version 1 | jq -r '.packages[].name' | sort
```

If the spec bench is imported, keep it as a diagnostics/benchmark tool. Do not
make normal mesh serving depend on it.