What made your custom research/task harness effective? How do you stop micromanaging?
retained text
I need to make my custom harness much better at **general research and other tasks (not coding)** that run longer than half a minute. Currently running: Qwen3.8 Flash Next FP8/GLM 5.3 Flash W4A16 Relevant capabilities already in my harness: * The system prompt is rendered in real time and can be changed based on almost any circumstance. * I have an effective skills system. I can trivially write instructions for any task. The agent will pick them up as needed. * The agent has access to web search, a Chromium browser instance, sandbox, etc. Problems: This might literally be a "skill issue" -- meaning I simply have to write better skills. But I feel like I am getting stuck micromanaging the agent, dictating every part of the process. I ask it to research something, it does two quick web searches and reports back some trivial garbage sales pablum. I poke it more and explain why that info is garbage, it does a little better next time. # How do you define when it's time to stop or what information is actionable? Must I really write a skill that says "come back with no less than 12 sources of this and this and this kind"? <-- This part is the question In other words, are you spending forever writing every single thing it has to do, or is there a much better way to leverage the intelligence of the model to make this call for you?
document fetched 2026-11-08 00:34 · canonical https://reddit.com/r/LocalLLaMA/comments/1xptj2q/what_made_your_custom_researchtask_harness