Skip to content
Conduit

skill

skill-tuff

Agent skill for building local, resumable AI pipelines.

unfamiliar-city/skill-tuffcommunitylistedunreviewed

Set up

Skills are installed by placing their folder where your agent loads skills.

The download comes from github.com/unfamiliar-city/skill-tuff itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 23, 2026

tuff

tuff-lil-unit

A Claude Code skill for building resumable AI pipelines with tuff-lil-unit.

Tell Claude what you want to build. Tuff handles the rest — crash recovery, concurrency, retry, and a queryable SQLite database of everything that happened.

An example pipeline

PhaseWhat it doesModelCallsConcurrency
CollectFetch 200 source URLsnone20050
ExtractPull structured data from each pageGPT-5-nano20010
ClassifyScore and categorise each resultGPT-5-mini2005
DistilAggregate into final reportnone11

~400 LLM calls, 200 HTTP fetches, 4 phases. One /tuff invocation.

Install

Via the unfamiliar-city marketplace:

/plugin marketplace add unfamiliar-city/marketplace
/reload-plugins
/plugin install tuff

Or add the skill directly to your project:

mkdir -p .claude/skills
git submodule add https://github.com/unfamiliar-city/skill-tuff .claude/skills/tuff

Use

Claude Code
Opus 4.6 · Claude API
~/Projects/myproject

 > /tuff build a pipeline that scrapes 200 URLs, extracts key data
   with GPT-5-nano, and generates a summary report

Claude designs the phases, sets up the file structure, installs tuff-lil-unit, and writes the code. If it crashes, re-run — it resumes from where it left off.

What the skill knows

  • Pipeline design — phases, table schemas, concurrency tuning
  • The full tuff API — ctx.step, ctx.upsert, ctx.model.*, ctx.agent.*
  • Three execution modes — LLM APIs (Anthropic/OpenAI), Claude Code headless (play at your own risk), any async function
  • Architecture patterns — file structure, fan-out, composition, progress tracking

Library

The skill installs tuff-lil-unit (npm i tuff-lil-unit) — a TypeScript toolkit that implements the step-function pattern with SQLite persistence. See the library repo for API docs and code examples.

Status

Alpha. See tuff-lil-unit for details.

License

MIT

↑↓ navigate openesc close