The README names one thing that sets the project apart, with evidence
A reader who has understood what the project does still has to decide against whatever they are already using. A README that never answers that leaves the comparison to them, and the alternative they already know wins by default. The evidence half is what stops the answer being an adjective: a boundary, a measurement, or a named competitor can be checked, and “powerful” cannot.
Observable check
README.md, before its first ## heading, presents at least one claim as what distinguishes the project from the alternatives a reader already has, that claim names a supported boundary, a measured number, or an alternative project by name, and it is traceable to the source, a manifest, the CI configuration, or a linked measurement. A number stated somewhere in the opening for another purpose does not satisfy this.
Where this comes from
- https://github.com/evilmartians/agent-skills/blob/main/skills/good-readme/SKILL.md
- https://evilmartians.com/chronicles/how-to-make-your-open-source-popular
Last read against these sources on 2026-07-28.
Same provenance and same caveat as R-DOC-01: a skill and an essay rather than a specification, weaker than the primary sources elsewhere in this file. This rule restores the third clause of upstream's opening block, what makes it different from alternatives, which R-DOC-01 dropped when it capped the opener at one sentence. It differs from upstream in two ways. Upstream puts the differentiator in the first paragraph; this rule accepts it anywhere above the first heading, which is where the facts list already lives. Upstream states it as a writing instruction; the evidence half, that the claim name a boundary, a number, or an alternative by name, is this standard's own, because a differentiator with no checkable noun in it is an adjective.