Tisty 1.16.0
The log moves to schema 14: a task now carries the layer you moved it to and
whether you opened it to agents. A machine still on 1.15 cannot read a store
this build has written to, and says so — «a newer Tisty updated your data,
update this one so the two agree again» — so update every machine you sync
before writing on any of them. Nothing else under the hood changed; the first
opening rebuilds the read cache once and is as quick as any other after that.
The terminal is yours, and the binary checks that it is
An assistant with a shell could type «tisty done 3» the day its MCP server was
not connected, and act as you; or «tisty agent --on», and let itself in. Every
command but «tisty mcp» now looks at who is at the keyboard before anything
opens — a mark in the environment, an assistant up the process tree, an editor
above it with no terminal — and turns a coding assistant away, saying which
mark did it so that you, taken for one by mistake, know what to change. Letting
an agent in from the terminal asks you, on the terminal; a shell with none is
refused. The task and document commands are frozen and retiring by stages: the
window is where Tisty happens, the binary stays as your assistant's door and for
maintenance. Settings no longer offers to put the command within reach; whoever
did before can still take it out.
A closed task is history to an agent
What you closed comes back to the assistant with a notice that it is done, or
dropped, and every tool that writes on it is refused: it reads as it ended, and
work that comes back is a new task that says how the last one went. Reopening a
task an agent said was done keeps its mark.
The archive reads in three layers, and you decide which
A closed task is a story or a trace by what it holds, unless you move it: «Move
to the trace» and «Move to the stories» in its detail, and the trail keeps each
move. A story is only hidden; a trace can be erased — one task at a time, from
its detail or its row — and the same rule holds in the terminal.
You can let an agent finish a task of your own
«Allow agents», in the detail of a task you wrote: the assistant may then say it
is done — you confirm — describe it, plan its steps and tick them as it goes.
Its day, its title, its list and its closing stay yours; «No agents» shuts the
door and keeps what was filled in. An agent that says a task is done has to
have ticked every step first, and it says done in the turn the work ends. Who
filed a task is signed by name in the row and the detail, «by canelo 39», and
the band that gathers what an agent said is done is called what it asks of you:
«To confirm».
The window holds a look of its own
Settings › Look: as the computer says, light, or dark.
Smaller
The footer of a task's detail fits on one line, with the rest behind «⋯». The
archive's controls lose their labels. A trace erased and filed again from the
same message points the agent at the later one. The read cache rebuilds itself
when the schema moves instead of falling silent for every upgrader. Erasing a
task keeps what it was written from, so an assistant reading the same message
again is told it was let go.
What's Changed
- feat: the terminal is the person's, and the binary checks that it is by @rgdevment in #42
Full Changelog: v1.15.1...v1.16.0