Detection of task running frameworks

[VeckoTheGecko]

How should brief interface with task running frameworks such as make, Pixi, Spin, and others? Perhaps on discovering a task running framework, the AI can be prompted to run the summary command for the available tasks (e.g., in Pixi's case pixi task list) or read the file directly?

[andrew]

brief already does some of this. It statically parses Makefile, justfile, Taskfile.yml and package.json and lists their targets under Scripts in the output, and those targets get linked to the Test/Lint/Build rows when the names match.

What it won't do is run the task runner to get the list. The Makefile parser deliberately avoids make -qp because that evaluates $(shell ...) and brief is often pointed at repos you haven't read yet. The same applies to pixi task list, invoke --list and spin, all of which load project code to enumerate tasks. So the line is: read config files, never execute.

For runners whose config is declarative that's fine. Pixi's [tasks] table in pixi.toml is plain TOML and could become another script source like the Taskfile one. Invoke's tasks.py and Spin's command modules are Python, so those stop at "here's the runner, go ask it yourself".

#74 adds tool entries for Just, Task, Pixi and Spin so they at least show up with their list command, which is the hint an agent needs to take it from there. #73 does the same for Invoke. Adding a pixi.toml script source is a reasonable follow-up if someone wants it.

[VeckoTheGecko]

So the line is: read config files, never execute.

I think that this makes a lot of sense.

From my POV I think #74 adding tool entries is really good and sufficient to close this

[andrew]

Merged 74, will get a new release out soon as well

[andrew]

Shipped in v0.7.0: https://github.com/git-pkgs/brief/releases/tag/v0.7.0