Fetching from the wire…
Public story · 2026-08-24 · high
The file requires strict TDD and a fresh coding session per feature, reloaded any time an agent's output quality slips.
Why now: The list stands out for spelling out literal rules instead of another philosophy post about coaching coding agents.
Programmer Fabien Sanglard published a working rulebook for keeping LLM-written code clean, in a file he calls agent.md.
Teams handing routine diffs to coding agents need some way to enforce standards a human reviewer would normally catch by eye. Sanglard turns that into rules an agent can check against before a diff goes out.
It isn't a philosophy doc. It's a checklist an agent has to satisfy before a diff goes out.
Extract magic numbers into named constants, and use early returns to cut nesting. Keep function names under 30 characters, use enums instead of boolean parameters, and brace every single-line conditional.
Layers only talk to their direct neighbors. Widening a method from private to public needs approval before an agent can make the change.
Testing follows strict TDD: the agent writes the test first, watches it fail, then writes the fix.
The context rule targets memory: one fresh session per feature, with agent.md reloaded mid-session the moment output quality drops.
Each link below shares sources, entities, or timing with this story.
Simon Willison released LLM / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (Simon Willison released LLM); both cover LLM, TDD; overlapping topics (agent, code).
Simon Willison released LLM / Shared entity: LLM / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Simon Willison released LLM); both cover LLM; overlapping topics (agent, alway).
Simon Willison released LLM / Shared entity: LLM / Shared topic / Earlier coverage
Linked by a graph relationship (Simon Willison released LLM); both cover LLM; overlapping topics (agent, approval).
Linked by a graph relationship (Simon Willison released LLM); both cover LLM; overlapping topics (agent, alway).
Linked by a graph relationship (Simon Willison released LLM); both cover LLM; overlapping topics (agent, approval).
Simon Willison released LLM / Shared entity: TDD / Shared topic / Earlier coverage
Linked by a graph relationship (Simon Willison released LLM); both cover TDD; overlapping topics (agent, code).
Linked by a graph relationship (Simon Willison released LLM); both cover TDD; overlapping topics (agent, code).
LLM uses OpenAI / Shared entity: LLM / Earlier coverage / Tension
Linked by a graph relationship (LLM uses OpenAI); both cover LLM; earlier LLM coverage from 2026-07-27.