weft-live-test
SkillDev toolsLets your agent walk you through running a node's live test against a real service with a real credential.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the weft-live-test skill
About this capability
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 sixteen `weft-` reference skills beside it are things you read; this is a procedure you carry out.
What this skill tells your AI
The instructions your AI receives, as published by weavemindai/weft in tangle/cline/.cline/skills/weft-live-test/SKILL.md and read by ahel’s review.
Run the live tier of a node's self-tests, with the user's informed consent. The node type or package is the one the user named when they asked. The local tiers (basic, fake) already passed when the node was built; the live tier calls the real service with a real credential and can spend money, which is why it is run here, by the user's choice, and not by the node specialist.
Walk it in this order:
- Say what it costs. One live test is one short-lived pod in the cluster making a real call on the named service, on the user's own credential or their credits. Get a plain "yes" before anything else.
- Find what the tests need. Read the node's
tests.rs(nodes/<snake_name>/tests.rs, the folder whosemetadata.jsondeclares that type, or the package's members) for theNodeTest::liveentries: the service each names, the credential fields, and any fixtures the test cannot self-provision (a chat id to message, a file to touch). - Check the ground.
weft daemon status; start it with the user if it is down. The live tier also needs the project registered with the dispatcher: if the run refuses with an unknown-project error, register with an ordinaryweft run(registration is part of it) or follow the error's own hint. - Set up the credential, never through the chat. Two ways, the user picks:
- a key: the user puts
WEFT_NODE_TEST_<SERVICE>_<FIELD>=...lines in the project's.env(auto-loaded) or their shell themselves. Never ask them to paste a secret into the conversation, and never echo one back. Fixtures go in the same file, under their own name:WEFT_NODE_TEST_<fixture name>=..., exactly as the test declares it (the missing-fixture error prints the variable it wants). - a connection they already made in the editor:
--connection <service>=<grant id>.
- a key: the user puts
- Run it.
weft test-node <type> --tier live(add--parallelif they want speed and several tests exist). The permission prompt and the CLI's own money confirmation are both expected; the user answers them. - Report. Each test line, pass or fail, in plain words; a failure carries the service's own error text. On green, note that
weft node-test-hash <type>prints the hash that lets a scripted sweep skip this package until its sources change, and that the node is now proven end to end.
Signals
- GitHub stars
- 2k
- Forks
- 221
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
weft-live-test- Source
- github.com/weavemindai/weft