mirror of
https://github.com/software-mansion/argent.git
synced 2026-09-14 19:27:14 +08:00
Ignacy Łątka
36af6ef5dc
The autofix-first step promised more than it delivers: "most of these rules
autofix, which usually leaves only a handful of findings needing judgment".
Measured against the ruleset this skill prescribes, one rule of fifteen has a
fixer `--fix` will apply.
Checked two ways. Rule metadata: only `react-native/no-single-element-style-
arrays` and `react-hooks/exhaustive-deps` declare `fixable`, and exhaustive-deps
exposes it as editor suggestions rather than something `--fix` applies. Then a
real run — the full ruleset over a component carrying inline styles, a colour
literal, an index key, arrow/object/array props and an empty dep array — went
from 13 problems to 11, and the only edit was unwrapping `style={[styles.one]}`.
Every high-volume rule survived: no-inline-styles, no-color-literals,
jsx-no-new-function-as-prop, exhaustive-deps.
That matters because the sentence is load-bearing in the new procedure: it tells
the agent the field will be mostly cleared before batching, so the four batches
read as small. They are not — they carry essentially the whole finding list, and
the cap is what keeps the work bounded, not autofix. Say that, and record the
fixability in the ruleset reference so it is checkable where the rules are.