Introducing ai-test-kit, a test library for AI SDK apps
I maintain a handful of packages in the AI ecosystem and keep repeating the same patterns:
Mocking a model, building parts, simulating a stream, ...
The kit abstracts that boilerplate into one-liners
Install in claude code:
/plugin marketplace add zirkelc/agent-plugins
/plugin install agent-journal@zirkelc
`aj sync init` needs the CLI from the shell installer
The README has both, plus the one-line `gh` command for a private repo:
https://t.co/RpHxgU5dVX
Git is already the best sync layer for an agent's notes
`agent-journal` can commit and push every entry claude code writes to your repo
Handy for remote or cloud agents: one place for all your entries
Off by default, and it only pushes to a remote you set
Introducing 𝚊𝚐𝚎𝚗𝚝-𝚓𝚘𝚞𝚛𝚗𝚊𝚕 plugin
Working with agents is fast-paced: you work on multiple projects in parallel, make decisions on every turn, generate ideas and build new things.
A week later you cannot answer simple questions like "what did we do last week?"
𝚊𝚐𝚎𝚗𝚝-𝚓𝚘𝚞𝚛𝚗𝚊𝚕 makes the model record its work along the way: one small markdown file per milestone, written on its own, in a format that's easy to read and scan.
A year later, the answer is one question to your agent, or one `grep`, away.
I tried append-only logs for the first iteration of agent-journal, but it has a few downsides IMO:
- a weaker model (way back with opus 4.5 and co) did accidentally overwrite instead of append
- permissions (this was before auto mode) are easier to grant for the Write() tool instead of a generated bash command that appends
- running multiple agents on different machines each with their journal, but all sync to one github repo (avoids merge conflicts)
@cnakazawa I can imagein the corporate politics around that decision :-D
Does Flow have any benefits now over TS, or do they use it just for historical reasons?