factory-llm-workflows

factory-llm-workflows is a skill for Claude Code from nonlinear-xyz/factory-kit. It costs 93 tokens per session (2,836 once invoked), scanned A, original, MIT.

A set of conventions for building LLM applications with LangGraph, a framework for connecting AI workflow steps, and FastAPI, a Python web framework. It covers shared workflow state, routing between steps, structured results, search with confidence checks, prompts, and streaming responses.

In plain words
What is it for?
Use it for chat systems, agents, retrieval-augmented generation (RAG), document question answering, structured data extraction, and streaming interfaces. RAG means finding relevant documents before asking the model to answer.
Why use it?
It prevents multi-step AI features from turning into loosely connected functions with unclear state and unreliable outputs. It gives each workflow step a defined input, output, and fallback behavior.

Skill for Claude Code

Written for Claude Code: shipped in a Claude Code plugin.

Part of the factory-kit plugin — 37 skills, 8 commands, 12 agents, 1 MCP server shipped together

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/nonlinear-xyz/factory-kit/factory-llm-workflows
Any agent
npx skills add nonlinear-xyz/factory-kit --skill factory-llm-workflows
Clone the repo
git clone --depth 1 https://github.com/nonlinear-xyz/factory-kit

Made for: Claude Code.

Or install factory-kit, the plugin that ships this one along with the rest of its 37 skills, 8 commands, 12 agents, 1 MCP server.

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 factory-llm-workflows

README.md
[![agentmods](https://agentmods.dev/badge/skills/nonlinear-xyz/factory-kit/factory-llm-workflows.svg)](https://agentmods.dev/skills/nonlinear-xyz/factory-kit/factory-llm-workflows)
Your own site
<a href="https://agentmods.dev/skills/nonlinear-xyz/factory-kit/factory-llm-workflows"><img src="https://agentmods.dev/badge/skills/nonlinear-xyz/factory-kit/factory-llm-workflows.svg" alt="Measured on agentmods" height="20"></a>
Per session 93 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,836 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00093 $0.02836
Opus 5 $0.00046 $0.01418
Sonnet 5 $0.00019 $0.00567
Haiku 4.5 $0.00009 $0.00284

Measured 6d ago against content hash 5da9ae65489d, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

factory-llm-workflows scanned grade A with 0 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 6d 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.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/factory-llm-workflows/SKILL.md · 256 lines

How it starts

The opening of the file, as written. The whole thing — 256 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Factory LLM workflows

Each section leads with Principle (one sentence, stack-agnostic), then Why (constraint → option → tradeoff), then Recipe (the LangGraph / FastAPI / SSE shape we use), and Failure mode when there's one to name. Sections that are pure style with no deeper truth are marked Recipe only.

State shape — TypedDict, not Pydantic

Principle. LangGraph state is a TypedDict, not Pydantic. The state library's merge semantics dictate the shape.

Why. LangGraph merges state between nodes by shallow dict update — the framework expects a dict-like object whose fields are independently updatable. Pydantic validates on construction; every partial update fails validation or requires .model_copy(update=...), which loses the simplicity. TypedDict matches the framework's semantics: it's a dict, fields are optional via total=False, the type annotations are documentation that the type checker enforces at call sites.

Recipe.

from typing import TypedDict
from typing_extensions import NotRequired

class ChatState(TypedDict, total=False):
    """Documented fields. total=False makes everything optional."""
    user_query: str
    intent: NotRequired[str]
    rewritten_query: NotRequired[str]
    retrieved_chunks: NotRequired[list[RetrievedChunk]]
    response: NotRequired[str]
    rag_fallback_attempted: NotRequired[bool]  # one-attempt loop guards

Nested TypedDicts (RetrievedChunk, EvidenceChunk, ClaimVerdict) for complex types.

Failure mode. Reaching for Pydantic state because "Pydantic is more rigorous" — every node update became a .model_copy(update=...) dance, and the graph wiring drowned in validation noise.

Graph composition — node factory closures

Principle. Nodes are produced by factory functions that close over their dependencies; the graph wires the result.

Why. A node that imports its dependencies (LLM client, prompt template, retriever) at module scope is hard to test and impossible to swap. A factory function takes dependencies as parameters and returns a callable; the graph passes the factory the wired-up dependencies. Testing is "construct the node with mocks"; swapping is "construct the node with the alternative."

Read the full file on GitHub · 256 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. 6d ago First seen · 256 lines · 93 tokens per session scan A 5da9ae65489d

Subscribe to this mod's changes

factory-llm-workflows is a skill published in the GitHub repository nonlinear-xyz/factory-kit (9 stars, last pushed 1mo ago), licensed MIT. It adds 93 tokens to every session and 2,836 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

hook-template

Generate hook script from template. Use when adding a new hook, wiring a PreToolUse/PostToolUse/Stop/Notification hook, or scaffolding hook config for settings.json.

claude-world/director-mode-lite · 39 tokens

agent-template

Generate custom agent from template. Use when creating a new subagent from scratch, or scaffolding an agent file with correct frontmatter.

claude-world/director-mode-lite · 30 tokens

workflow

Run the complete 5-step development workflow: focus problem → prevent over-development → test-first (TDD) → document → smart commit. Use when starting a new feature, or when the user runs /workflow or asks for the full development flow.

claude-world/director-mode-lite · 52 tokens

check-environment

Verify Claude, Codex, and Grok availability plus Director guidance, relay, agents, and skills. Audit optional hooks only when selected. Use after installation or when a native surface misbehaves.

claude-world/director-mode-lite · 44 tokens

doc-writer

Documentation templates and standards: README structure, API reference format, changelog (Keep a Changelog), and comment guidelines. Use when creating or updating documentation. Loaded automatically by the doc-writer agent.

claude-world/director-mode-lite · 44 tokens

smart-commit

Create clean Conventional Commits: inspect the diff, group related changes, run quality checks, and write type(scope) messages. Use when committing work, or when the user runs /smart-commit or asks to commit changes.

claude-world/director-mode-lite · 49 tokens