Your repo has 5 instruction files but your coding agent only reads 1
bibryam · x · 2026-11-10
Luca Mezzalira explains that most instructions you write for coding agents are never read: at session start only the root AGENTS.md loads in full, nested AGENTS.md files load only when the agent works in that folder, and skills expose just their name and description until the agent decides to open them.
After encoding 10 years of micro-frontend decisions into skills, he found the same prompt yields very different results with and without them — but a skill only matters once the agent opens it, based solely on its name and description.
His advice: keep never-break rules in the root file, and spend as much effort on a skill's description as on its body.
More from coding & agent
- AI builds a 3D scene in real time: 120s 1080p60 capture done in 122s wall time, minor Z-fighting quirks — teortaxesTex · 2026-11-10
- NODES 2026: 100+ sessions, but devs should filter for Text2Cypher, entity resolution and quantization — JeremyCMorgan · 2026-11-10
- A One-Person AI Company in Hong Kong Staffed by a Team of Grok Bots — AIFlow_ML · 2026-11-10
- Astral open-sources serc under new astral-experiments org for agent-authored projects — charliermarsh · 2026-11-10
- Astral's serc: Codex-built Rust CPython bytecode compiler, byte-identical and 2-4x faster — charliermarsh · 2026-11-10
- Sentry CEO shares hands-free AI game dev progress: auto-added wildlife shaders at 82fps — zeeg · 2026-11-10