Queueing multiple skills at once

I’m trying to keep a project at 100% coverage.

It seems that telling Cursor to get to 100% gets interpreted as a “Big Task” and spooks Cursor. The quality of the reasoning goes down, even when there are only a few files that need testing.

I have a tool that analyzes which files still need test coverage adding. And I have a skill to tell Cursor “add test coverage to this file”, which works reliably.

It is possible for me to enter multiple commands into a chat window, and have them executed one by one? e.g. for the commands below. I couldn’t figure out how to add them at once, so had to add them one-by-one, which doesn’t feel great.

/improve_test_coverage src/Bristolian/AppController/Admin.php
/improve_test_coverage src/Bristolian/CliController/BccTroFetcherCliController.php
/improve_test_coverage src/Bristolian/CliController/GenerateFiles.php
/improve_test_coverage src/Bristolian/Parameters/PropertyType/OptionalRoomContentListOrder.php
/improve_test_coverage src/Bristolian/Repo/RoomRepo/FakeRoomRepo.php
/improve_test_coverage src/Bristolian/Repo/UserRepo/PdoUserRepo.php
/improve_test_coverage src/Bristolian/Repo/WhatDoTheyKnowRequestEventRepo/PdoWhatDoTheyKnowRequestEventRepo.php
/improve_test_coverage src/Bristolian/Service/AvatarImageStorage/FakeAvatarImageStorage.php
/improve_test_coverage src/Bristolian/Service/WhatDoTheyKnowFeedFetcher/FakeWhatDoTheyKnowFeedFetcherReturningJson.php
/improve_test_coverage src/functions.php
/improve_test_coverage src/functions_whatdotheyknow.php

Edit - oh this is solved by sub-agents isn’t it. Not going to delete, this but am looking them up.

Word it differently, don’t ask it to do full coverage, say I need tests these files 1, 2,3 .. that you know are missing coverage.

Hey, thanks for the request. You already figured it out in your edit. Yep, that’s exactly what subagents are for. There’s no chat feature to paste a bunch of commands and run them one after another from a single message, but there are two options that solve it:

  1. Async subagents, I’d recommend this for your case. In Cursor 2.5, subagents became async. The main agent can run many subagents in parallel, each in its own context. For running /improve_test_coverage across N files, this fits perfectly. Ask the agent to dispatch subagents for the list of files. Your main context stays clean, and they work independently. Details: Subagents | Cursor Docs and Cursor 2.5: Async Subagents.

  2. A regular message queue. While the agent is working, you can send the next prompt and it’ll go into the queue and run after the current one. Downside is you still need to hit send N times, but it’s guaranteed to run sequentially.

For 100% coverage, I’d try subagents. Use one prompt like: “here’s the list of files without coverage, dispatch one subagent per file and run my /improve_test_coverage skill there” plus your list. Let me know how it goes.