// Use this skill when inspecting GGUF models, planning layer ranges, generating or validating skippy package artifacts, fake packages for direct GGUFs, materialized stage cache behavior, or GGUF writer integration.
Use this skill when running benchmark orchestration, local single-stage or split benchmarks, benchmark report flow, or performance-oriented skippy runtime checks.
Use this skill when validating skippy staged execution against full-model execution, adding model families, changing split boundaries, testing activation wire dtypes, or diagnosing mismatch behavior.
Use this skill when running, configuring, debugging, or embedding skippy-server, binary stage transport, OpenAI frontend integration, activation wire dtype settings, stage configs, lifecycle status, or nonblocking telemetry.
Use this skill when certifying mesh-llm KV/cache stability under repeated OpenAI tool-call loops, same-prefix cache reuse, suffix-prefill limits, or native Skippy slot/decode/eviction failures.
Use when changing mesh-llm automation or CLI flows that discover Hugging Face GGUF models, plan CPU Hugging Face Jobs for layer-package splitting, estimate max cost, or publish skippy layer packages/catalog entries.
Use this skill when adding, renaming, removing, or reviewing mesh-llm OTLP metrics, telemetry attributes, metrics exporter settings, or telemetry documentation.
| name | skippy-model-package |
| description | Use this skill when inspecting GGUF models, planning layer ranges, generating or validating skippy package artifacts, fake packages for direct GGUFs, materialized stage cache behavior, or GGUF writer integration. |
| metadata | {"short-description":"Inspect and package GGUF stages"} |
Use this skill for model inspection, package planning, stage materialization, and cache behavior.
Rust owns package manifests, topology planning inputs, cache policy, and mesh model-storage integration. The patched llama/skippy ABI owns GGUF tensor inspection and GGUF artifact writing.
Direct GGUF loading in mesh should materialize as a fake package identity in the skippy runtime so the split-serving path can use the same package-backed stage machinery as Hugging Face packages.
Check current package names before running commands:
cargo metadata --no-deps --format-version 1 | jq -r '.packages[].name' | sort
Useful current checks in this repo:
cargo test -p skippy-runtime --lib
cargo test -p skippy-topology --lib
cargo test -p mesh-llm-host-runtime --lib inference::skippy
For a published layer package, prefer package-local diagnostics before a live split smoke:
cargo test -p skippy-model-package --bin skippy-model-package
skippy-model-package preflight <package-dir> --stages 2
Materialized stages are derived cache. Model storage commands may evict materialized stage artifacts without deleting the source model/package. Preserve pinned materialized artifacts unless the command explicitly asks for a stronger cleanup.