Tip no. 7, teach your agent to write better: upgrade old models first. My skills: one flags AI clichés, one is trained on my own writing and /tldr cuts answers to 3 points. See: https://t.co/sYSD9fgbPc
Tip no. 6, use agents to filter what reaches you. Mine builds a daily feed from blogs and social posts that I read on an e-ink tablet, no ads, no irrelevant posts. And a skill reads my config and picks out the Claude Code changes relevant to me: https://t.co/IvSPQOryym
Tip no. 5, practice agentic engineering: I built a photo editing app over the weekend without writing a line of code, yet with a lot of software engineering. Not vibe coded, I care about the code. Analysis: https://t.co/LP2dx8fbUi
Tip no. 4, make stories from PR code reviews: review from a higher level by grouping changes into areas, and give the agent targeted feedback right in the code. Combining both is not that hard: https://t.co/m0JQVMtY5W
We had a great time at @PragVue_conf. This year with a hackathon integrated into the conference! We have built an actual project there: https://t.co/F0bf9bgmOp
Tip no. 3, save 1/3 of the initial context in every Claude session: each tool and feature costs tokens at startup. Turn off what you rarely use in settings.json and stay in the smart zone longer. Any feature can come back per session. My setup: https://t.co/6rClt2UKfZ
Tip no. 2, grill until shared understanding: before the agent writes a plan, make it interview you, one decision at a time. Left alone it fills gaps silently and its guesses look like your choices. Concept: https://t.co/YkK4dYzXUo
Tip no. 1, one-off tools: don't hesitate to build a tool or a whole program for a single job. A woodworker makes one when nothing at hand fits. It adds precision and elegance, and with an agent it takes minutes. Demo: https://t.co/0FAJboBSs3
@mattpocockuk I am using kitty terminal and its programmable capabilities are enough, no need for other tools. I am using custom /split command to create a fork of current session in side by window.
@mattpocockuk I have a skill that takes tickets and askes Claude to implement them one by one in subagents. The tickets are prepared with your to-tickets skill. I don't see anything missing. Using Fable as much as my plan allows though.
@zeeg I agree that specs are hard to be treated as docs. But I do save my specs and other artefacts in chronological order in a folder in every project and I am happy with that setup for months. The usefullness is temporary though.