docs: add view queries and content projection reference to developer skill - #70133

@aminesbdev thanks for sending this PR. I have some questions for you about it:

1. Have you noticed that agents are unable to generate view query content and work with content projection?
2. Have you been able to measure or track the improvements to agentic coding with some eval tool, for example, https://github.com/mgechev/skillgrade?

@aminesbdev thanks for sending this PR. I have some questions for you about it:

1. Have you noticed that agents are unable to generate view query content and work with content projection?
2. Have you been able to measure or track the improvements to agentic coding with some eval tool, for example, https://github.com/mgechev/skillgrade?

Thanks for the review @MarkTechson

Regarding your first point, yeah, I've noticed coding agents struggle a lot here. Since they're mostly trained on older codebases, they default to @ViewChild and @ContentChild. When they do try using signal queries, they often mess up the syntax (like trying to pass { static: true } which doesn't compile). They also don't know about Angular 18's new fallback content in , so they end up writing a lot of custom boilerplate to work around it.

On skillgrade, I haven't set up any evals yet. Since we don't have a testing harness in the repo for these skills, I wanted to get the content right first. But I'm down to write an eval.yaml and test it out if you think that's a good next step.

@MarkTechson
I ran the evaluation locally over 10 trials for each task using gpt-4o model (with a deterministic grader verifying the migration of @ViewChild('myInput', {static: true}) to viewChild.required('myInput')).

Here is some screenshot of my local skillgrade HTML preview showing the comparative results

Even the state-of-the-art gpt-5.6-sol reasoning model has a 10% failure rate due to inconsistent formatting and legacy options fallback (90% Pass Rate over 10 trials).