A programming language for AI orchestrations
This repository also configures its own agents. See what weft tells them →
Skill Claude CodeCodex
COMMAND, not reference: Walk the user through setting up and running a node's live test. Run this when the user asks for this step by name, optionally naming the node type or package name. The weft- reference skills beside it are things you read; this is a procedure you carry out.
Skill Claude CodeCodex
COMMAND, not reference: Expert entry: have a specific node built, or take the hand and write it yourself. Run this when the user asks for this step by name, optionally naming what the node should do, in plain words. The weft- reference skills beside it are things you read; this is a procedure you carry out.
Skill Claude CodeCodex
COMMAND, not reference: Build and run the weft program, then report what came out. Run this when the user asks for this step by name, optionally naming node ids to target. The weft- reference skills beside it are things you read; this is a procedure you carry out.
Skill Claude Code
Read when a program serves several PEOPLE, each with their own account, number, bot, spreadsheet or schedule (a WhatsApp assistant for many customers, a Slack bot per workspace, a report per client): the table that says which person owns which instances, one instance per person as the simple case, people who own…
Skill Claude Code
Read when the user wants weft itself on their Google Cloud for the first time, or to upgrade or resize the weft already there: getting the user's GitHub and Google Cloud accounts and CLIs ready, the fork, the one-time gcloud block, the database, the fork's variables and secrets, running the install workflow, the first…
Skill Claude Code
Read when the user wants weft or a program live for real people on their cloud, asks about prod, a target, --on, a domain or GCP, or a deploy failed: targets and logging in, domains, the deploy workflow weft ci add writes, handing it its settings with weft target export, a project's frontends (weft frontend), what a…
Skill Claude Code
Read whenever part of a program must run as several separate copies, each under its own id, even with ONE user: a session per chat, a sandbox per job, a workspace per document, a bot per customer. @perinstance (a container each), @instancefilled (a value each, a connection included), which instance a run is for, the…
Skill Claude CodeCodex
The only agent that acts on a cloud install. Dispatched by Tangle with what to do on a named target (deploy, roll back, set up CI, read a failed cloud build); names the target with --on on every command, reads the weft-deploying skill, and reports the deployed version and route addresses, or the exact failure.
Skill Claude CodeCodex
COMMAND, not reference: Deploy the program to a cloud install, through the deployer. Run this when the user asks for this step by name, naming the target and optionally what to do. The weft- reference skills beside it are things you read; this is a procedure you carry out.
Skill Claude Code
Read when a cloud install needs its database: the user has no Postgres for it, or asks which to use. Gets one that scales to zero (Neon), with the user only signing up and pasting one key, and puts its addresses in the fork's secrets. Part of weft-cloud-install.
Skill Claude Code
Read when a run failed, did the wrong thing, never started, got stuck or was cancelled, or when you need to read what a past run did: the order to read a run in (executions, logs, events), what each skip and cancel reason means, and the debugging playbook, from a compile error to a daemon that does not answer.
Skill Claude Code
Read when a program asks a person something or a person starts a run (HumanQuery, HumanTrigger), or the user asks about the browser extension or a task token: how a waiting question reaches people, building, loading and connecting the extension, minting and scoping a token, and what a waiting run costs.
Skill Claude Code
Read when a trigger needs the internet to reach weft (a route, a provider that pushes its events), when the user asks about a public URL or a tunnel, or when activation fails with "needs your weft to be reachable from the internet": what the error means, walking the user through ./setup.sh --public-url, the dial-out…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: