Jev read every CLAUDE.md
What 58 CLAUDE.md files actually say to the agent.
Claude Code started reading AGENTS.md last week. GitHub now has over half a million public CLAUDE.md files. TypeSafe's Jev, a decision model that returns probabilities instead of prose, read a size-stratified sample of them and answered eleven questions about each. Then it can read yours.
58
CLAUDE.md files read
sampled from 0 public ones
45%
start with the /init boilerplate
14% look never edited
47%
tell the agent to verify its work
71% give exact commands
31%
contain a hard “never”
0% shout in caps
7%
read as LLM-written
0% leak something private
What files contain
Share of files where Jev put the probability above 50%.
- Exact build / test / lint commands71%
- Tells the agent to verify its work47%
- Maps where things live in the codebase78%
- Explicit prohibitions (never, do not)31%
- Git rules: commits, branches, pushing2%
- Bans comments or explanatory prose2%
- Reads as LLM-generated7%
- Shouts in ALL CAPS0%
- Leaks something that should be private0%
What kind of document it is
- Architecture map45%
- Ops manual41%
- Style guide9%
- Persona5%
Does length help?
| size | files | commands | verify | /init |
|---|---|---|---|---|
| <1k | 2 | 0% | 0% | 0% |
| 1-2.5k | 6 | 83% | 33% | 17% |
| 2.5-5k | 19 | 84% | 63% | 68% |
| 5-10k | 18 | 72% | 50% | 39% |
| >10k | 13 | 54% | 31% | 38% |
Grade yours against the 58
one Jev call, about a tenth of a cent, on your own keyFree at console.typesafe.ai/keys, $5 credit included. The key is forwarded per request and never stored or logged.