mesh-llm: Skill for Claude Code

.agents/skills/llama-stage-patch-changes/SKILL.md

llama-stage-patch-changes is a skill for Claude Code, Codex from Mesh-LLM/mesh-llm. It costs 51 tokens per session (2,138 once invoked), scanned C, original, Apache-2.0.

A set of rules for changing mesh-llm’s modified llama.cpp runtime interface, native hooks, model handling, tensor filtering, activation-frame execution, and GGUF writing. llama.cpp is software used to run large language models, and GGUF is a model file format.

In plain words
What is it for?
Use it when changing the patched llama.cpp code, its internal interface, model lifecycle, loading, package creation, or the Rust code that connects to it.
Why use it?
It keeps native runtime changes organized, reviewable, and matched with the Rust code that calls them. It also helps diagnose mismatches between the two sides.

Skill for Claude CodeCodex

Written for no agent in particular: nothing here depends on one. Also seen: installed under .agents/ (shared by several agents).

This is Mesh-LLM/mesh-llm's own configuration. It tells Claude Code and Codex how to work on mesh-llm itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything mesh-llm configures →

Needs its repository: it runs a file that does not travel with it, so clone the repository first. The line is python3 scripts/generate-skippy-api-doc.py.

About the project

Mesh LLM is a distributed AI system that combines GPUs and memory from multiple machines and exposes them through one OpenAI-compatible inference server, which is a service that runs models and answers API requests. It is for people who want to share compute privately or publicly to run models for agents and chat, including models too large for one machine. The catalogue add-ons support workflows for operating and using the mesh.

Mesh-LLM/mesh-llm · 3,358 stars · on GitHub · meshllm.cloud

Reuse

Borrowing it

Nothing to install: this file belongs to Mesh-LLM/mesh-llm. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/Mesh-LLM/mesh-llm/main/.agents/skills/llama-stage-patch-changes/SKILL.md
Clone the repo
git clone --depth 1 https://github.com/Mesh-LLM/mesh-llm

Made for: Claude Code, Codex.

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 llama-stage-patch-changes

README.md
[![agentmods](https://agentmods.dev/badge/skills/mesh-llm/mesh-llm/llama-stage-patch-changes.svg)](https://agentmods.dev/skills/mesh-llm/mesh-llm/llama-stage-patch-changes)
Your own site
<a href="https://agentmods.dev/skills/mesh-llm/mesh-llm/llama-stage-patch-changes"><img src="https://agentmods.dev/badge/skills/mesh-llm/mesh-llm/llama-stage-patch-changes.svg" alt="Measured on agentmods" height="20"></a>
Per session 51 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,138 The whole file, excluding the scripts and references it only reads on demand.
Security scan C 1 finding. A grade says what 26 rules found in the file — not that it is safe.
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 $0.00051 $0.02138
Opus 5 $0.00026 $0.01069
Sonnet 5 $0.00010 $0.00428
Haiku 4.5 $0.00005 $0.00214

Measured 7d ago against content hash 2fb689d98880, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade C, and why

llama-stage-patch-changes scanned grade C with 1 finding 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 7d 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.

Recursive force deletehighDestructive command

rm -rf with a variable or a broad path is one typo away from removing the wrong tree.

trap 'rm -rf -- "$tmp_root"' EXIT
.agents/skills/llama-stage-patch-changes/SKILL.md · 204 lines

How it starts

The opening of the file, as written. The whole thing — 204 lines — stays where its author put it; the contents beside it link to each section on GitHub.

llama-stage-patch-changes

Use this skill when changing the Skippy staged-runtime ABI carried in third_party/llama.cpp/patches.

Boundaries

  • Keep durable llama.cpp-side changes in third_party/llama.cpp/patches/*.patch.
  • Keep the upstream pin in third_party/llama.cpp/upstream.txt.
  • Do not edit .deps/llama.cpp as the final artifact; regenerate the patch queue from commits.
  • Keep mesh orchestration, protocol compatibility, lifecycle, model management, and API status behavior in Rust.
  • Keep one functional boundary per patch. Patch numbers must be unique and contiguous.
  • Keep public ABI declarations separate from independently reviewable model lifecycle, loading, and package implementation changes.
  • The Skippy native ABI is an internal lockstep boundary, not a stable cross-version compatibility contract. It may change whenever the feature requires it; update the Rust FFI mirror and all callers in the same change.
  • Do not preserve old native ABI signatures for compatibility. Bump the ABI version when the boundary changes so mismatches are diagnosable, and make sure the shipped Rust side and native runtime are built from the same queue.
  • Do not add a terminal source-reorganization patch. A deliberate layout or ownership change must be represented in the recreated patches that own the affected capabilities.

Native Source Layout

  • include/skippy.h is an umbrella only. Put public C ABI declarations in standalone include/skippy/<capability>.h headers.
  • Put implementations in src/skippy/<capability>.cpp and private C++ declarations in narrowly named src/skippy/*.h headers.
  • Use snake_case capability names. Keep exported symbols prefixed with skippy_ and avoid generic helpers, utils, or expanded common modules.
  • src/skippy.cpp is retired. Extend the owning capability module and keep new implementation files below 1,000 lines.
  • Make every public header independently compilable as both C11 and C++17. Update explicit CMake source lists and installation rules with new modules.
  • Do not preserve retired source include paths unless the task explicitly asks for compatibility. Continue to version and mirror any binary ABI change.

Read the full file on GitHub · 204 lines

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. 7d ago First seen · 204 lines · 51 tokens per session scan C 2fb689d98880

Subscribe to this mod's changes

llama-stage-patch-changes is a skill published in the GitHub repository Mesh-LLM/mesh-llm (3,358 stars, last pushed today), licensed Apache-2.0. It adds 51 tokens to every session and 2,138 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

developing-genkit-tooling

Best practices for authoring Genkit tooling, including CLI commands and MCP server tools. Covers naming conventions, architectural patterns, and consistency guidelines.

genkit-ai/genkit · 35 tokens

architecture-decision-record

ADR templates in the Nygard format with context, decision, consequences, and alternatives. Use when writing ADRs, recording an architectural decision, or evaluating options.

yonatangross/orchestkit · 38 tokens

zcfg

Integrate zcfg (Zero Dependency Configuration Utility) into Java applications. Use when adding configuration loading, reading properties files, setting up application configuration, or integrating zcfg into a Java project. Triggers on "zcfg", "add configuration", "load properties", "application configuration with…

AdamBien/airails · 75 tokens

data-structures

Implements custom JavaScript data structures: queues, deques, stacks, linked lists, cons lists, circular buffers, unrolled lists, tries, heaps, graphs, LRU caches, CRDTs, pools, structs. Use when building or choosing non-native collections, optimizing enqueue/dequeue, designing persistent lists, or when the user asks…

metarhia/metaskills · 84 tokens

microprofile-server

Architecture and coding rules for long-running Java MicroProfile / Jakarta EE server applications — BCE layering, business components (BC), JAX-RS resources, CDI, JSON-P, testing (unit/integration/system), and Maven project structure. Use when creating, generating, scaffolding, writing, or reviewing code, resources…

AdamBien/airails · 87 tokens

bce

Generic, composable architecture rules for the Boundary-Control-Entity (BCE/ECB) pattern — business components, layer responsibilities, package structure, and cross-component relationships. Technology-neutral; meant to be composed with language- or framework-specific skills (e.g. microprofile-server, web-components…

AdamBien/airails · 151 tokens