I rarely code in the web interface, are you using Claude Code? Your first AGENTS.md is looks okay, but a good suite of skills is great (the trick with those is to deploy them at the right time… use AGENTS.md to identify that right time, so that you’re not always burning context). The other key thing is having lots of small, heirarchal docs files throughout the codebase, so the agent can navigate and learn what it needs to at each folder level.
I never said anything about web interface? But I am using a mix of Cursor, Codex, and Claude Code.
I don’t really trust skills, do they work well from your experience? I never find myself to have the need for skills when a single prompt file works well enough, it is not like it is that many tokens (3.5k for my prompt). And most of the skills look heavily LLM generated and complete garbage.
Small, hierarchical docs files could be useful as a token saving measure, but I mostly care about steering the model to do the correct thing. It is fine if the model has to read the code, module/function level docstrings work well enough.
I had a brief look at the Nori skills, so far not too impressed
I rarely code in the web interface, are you using Claude Code? Your first AGENTS.md is looks okay, but a good suite of skills is great (the trick with those is to deploy them at the right time… use AGENTS.md to identify that right time, so that you’re not always burning context). The other key thing is having lots of small, heirarchal docs files throughout the codebase, so the agent can navigate and learn what it needs to at each folder level.
I mostly just use Nori:
https://tilework.tech/
The author’s a regular on e.g. ACX, and his own blog has a lot of great takes and practical tips:
https://12gramsofcarbon.com/t/ai
I never said anything about web interface? But I am using a mix of Cursor, Codex, and Claude Code.
I don’t really trust skills, do they work well from your experience? I never find myself to have the need for skills when a single prompt file works well enough, it is not like it is that many tokens (3.5k for my prompt). And most of the skills look heavily LLM generated and complete garbage.
Small, hierarchical docs files could be useful as a token saving measure, but I mostly care about steering the model to do the correct thing. It is fine if the model has to read the code, module/function level docstrings work well enough.
I had a brief look at the Nori skills, so far not too impressed