@burkeholland I've setup the adversarial review skill you've talked about, and integrated into your orchestrator skill. The only problem is I quickly get rate limited on the number of Copilot model requests, once the review starts.
Any advice?
@burkeholland Yes, the skill did malfunction. It spawned hundreds of agents and that's what caused the rate limiting. I'm re-running now and so far so good.
@burkeholland I installed skill creator from skills.sh (the top one, from anthropics). Then I used your same prompt to create the adversarial review.
It's hard (for me) to say for sure how many agents; at least one each of:
- sonnet 4.6
- opus 4.6
- gpt-5.4
- gpt-5.3-codex
Could two followers please copy and re-post this tweet? I'm trying to demonstrate that someone is always there, especially at this time of year. Call 800-273-8255 (National Suicide Prevention Helpline US). Just two. Any two. Copy, not retweet. Let’s all look out for each other.
I'm sure you have probably seen this video but it brings a smile to my face and is a brutally honest assessment of how testing in many situations can fail. You know what I mean, fix code from one repository but the QA testers are retrieving from another repo.