LLM Tools

Use when a session is about to bulk-read 3+ files or any single file over ~400 lines, generate predictable boilerplate (tests, fixtures, config scaffolds, doc templates), or compress a long session transcript — and a cheap OpenAI-compatible worker model (Kimi, DeepSeek, OpenRouter, local Ollama) is available via the llm-ask/llm-write/llm-extract CLIs.

37 stars 7 forksAI, LLMs & Agents

Install

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add iamneilroberts/claude-skills --skill llm-tools
Manual install
git clone --depth 1 https://github.com/iamneilroberts/claude-skills.git /tmp/claude-skills
mkdir -p ~/.claude/skills
cp -r /tmp/claude-skills/skills/llm-tools ~/.claude/skills/llm-tools

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# llm-tools — cheap-worker delegation

## Overview

Premium-model context is the scarcest resource in a session. These three CLIs route
bulk I/O and predictable generation to a cheap long-context worker; the premium model
consumes the short answer instead of the raw files. Install + provider config:
[README.md](README.md) in this directory (~1-minute setup, any
`/v1/chat/completions` provider).

## Routing rules (the rules save the tokens; the CLIs just execute)

| Situation | Do this |
|---|---|
| Would read 3+ files, or one file >~400 lines, to answer a question | `llm-ask <files...> -q "..."` — read the answer, not the files |
| Need tests / fixtures / config scaffold / doc template from a spec | `llm-write -r <ref> -s "<spec>" -o <path>` — then review and edit; never trust blind |
| Long transcript needs to become doc updates | `llm-extract <transcript.jsonl> --sections summary,decisions,todos` |
| A mechanical step repeats ≥3× identically (same transform over many items) | Write a throwaway `scripts/<name>.sh` and run *that* — automate the repetition in code; don't spend model tokens re-prompting the same operation |

## Never delegate

- Architectural or design decisions
- Debugging — cheap models miss subtle bugs
- Anything touching auth, payments, PII, deletion, or production data
- Final commits and PR descriptions

If a delegated answer looks wrong or thin, re-read the sources yourself rather than
Read the full skill on GitHub

Details

Skill name
llm-tools
Path
skills/llm-tools/SKILL.md
Repo updated
21 August 2026

FAQ

What does the LLM Tools skill do?

Use when a session is about to bulk-read 3+ files or any single file over ~400 lines, generate predictable boilerplate (tests, fixtures, config scaffolds, doc templates), or compress a long session transcript — and a cheap OpenAI-compatible worker model (Kimi, DeepSeek, OpenRouter, local Ollama) is available via the llm-ask/llm-write/llm-extract CLIs.

How do I install the LLM Tools skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add iamneilroberts/claude-skills --skill llm-tools. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is LLM Tools popular?

It lives in iamneilroberts/claude-skills, which has 37 GitHub stars and 7 forks.

Related skills

Brag

37

iamneilroberts/claude-skills

Memorialize something you just built by generating a self-contained "showcase" page for it — a persuasive, good-looking deep dive: what it is, why it's impressive, how it works — with a screenshot if one can be found or captured, and links to related showcase pages. This is the sales/hype-leaning counterpart to /write-post (which drafts a neutral, no-BS technical post); reach for /brag when you want it to sell. Fast by design, like /idea: ask only what's needed, build the page, don't turn it into a project. Triggers on `/brag <subject>`, "brag about this", "make a showcase page for X", "write

Writing & Content

Branch

37

iamneilroberts/claude-skills

Create an isolated git worktree for a task and log it to a shared work journal, so multiple Claude Code sessions run in parallel without racing on HEAD, uncommitted edits, or the index. Subcommands: `list`, `done <slug>`, `update` (heartbeat), `coord <message>` / `coord clear <slug>` (cross-session constraints). Triggers on `/branch`, "start in a worktree", "isolate this work", "parallel session", "heartbeat the journal".

Git & Code Review

Claude Code Best Practices

37

iamneilroberts/claude-skills

Use when asked about Claude Code workflow, prompt caching cost, session economics, CLAUDE.md hygiene, context-window management, plan mode, permission modes, hooks, skills, subagents, checkpointing/rewind, parallel sessions, worktrees, headless/CI usage, the Agent SDK, slash commands, MCP configuration, or "what's the right way to use Claude Code for X". Also use proactively BEFORE suggesting /model, /compact, fast-mode toggle, or editing CLAUDE.md mid-session, since those bust the system cache.

AI, LLMs & Agents

Review

37

iamneilroberts/claude-skills

The top-level review verb — decide WHETHER, WHEN, WHICH, and HOW HARD to review a change before reaching for a review tool. Assesses the situation (a plan vs a diff, the risk surface, what changed since the last review, rounds already spent), picks the lightest review shape the risk justifies (in-session self-review → cross-model → multi-model panel), sets an initial round budget it will extend only on convergence, and dispatches. Exists to stop review spirals: reviewing before the plan is formed, reviewing after every trivial edit, or looping a reviewer past the point of convergence. Triggers

AI, LLMs & Agents

Gemini Review

37

iamneilroberts/claude-skills

Get a second-opinion review of the current diff (or a chosen focus — architecture, security, docs, design, performance, tests) from the Gemini CLI, then present its findings alongside your own analysis. Single-reviewer counterpart to /codex-review and /review-panel; needs the `gemini` CLI installed. Triggers on `/gemini-review [focus]`, "gemini review this", "get a gemini second opinion on this diff".

AI, LLMs & Agents

Review Panel

37

iamneilroberts/claude-skills

Run multiple independent code reviewers (Codex + Gemini + a fresh Claude) on the current diff or a plan, merge their structured verdicts, and apply a consensus-gated challenge round — returning a pass/fail exit code. Use for "panel review this", "get a multi-model review before I ship", "second-opinion review this diff", or /review-panel. Complements single-reviewer /codex-review.

AI, LLMs & Agents

Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.