What we can check, and what we can't
A wrong "verified" is worse than no answer. When the checker can't read something, it says so and leaves it out of the verdict.
What the checker verifies
- Cast-on and bind-off lines, including per-size counts written as CO 96 (104, 112, 120) and a clause in front of them such as "Using smaller needles, cast on 72 sts".
- Row and round instructions built from k, p, k2tog, ssk, yo, m1L, m1R, sl, the twisted k1tbl and p1tbl, and the wider decreases (skp, sl1-k1-psso, k3tog, cdd, s2kp), in their usual spellings.
- Bracketed and starred repeats, including a bracketed repeat nested inside a starred one, which is what Feather and Fan is made of, and counts spelled out as "twice" or "three times".
- Repeats written with parentheses, such as "(K2, p2) 9 times", beside per-size values on the same line: stitches in parentheses with a count after them read as a repeat, and bare numbers in parentheses after a value keep reading as sizes.
- Rows filled to the end however you say it: "to end", "around", "across", "to last 2 sts", by any stitch and not only knit and purl.
- A bind-off partway through a row, such as "Bind off 8 (9, 10) sts, purl to end", with or without a row header in front of it, which is how most flat garments shape a shoulder or an armhole.
- Short rows in their common forms: a row ending in "w&t", "wrap and turn" or "turn", heel turns that decrease before turning, "Short row 1:" headers, and the pick-up row written "Knit to end, picking up wraps". The stitch count is carried across the section instead of lost.
- Shaping phrases such as Dec 1 st each side every RS row 3 (4, 4, 5) times, expanded to the rows they cover.
- Whether a stitch pattern still lines up after a section changes the stitch count.
- Row numbering: a number that repeats, a gap where one is missing, and two rows in a row marked the same side.
What it can't check yet
- Crochet. Stitchwright is knitting-only for now, and says so rather than guessing at sc and dc.
- Charts pasted as images. A chart has to be derivable from rows the checker could read.
- Stitches outside the list above: cables (C4F, C6B), twisted decreases (k2tog tbl), a psso spelled across commas ("sl 1, k1, psso"), German short-row double stitches (DS), and a turn in the middle of an instruction list ("k5, turn, p5"). A row holding one is named as unread rather than counted.
- Bind-off shaping written as prose, such as "Bind off 4 sts at beg of next 2 rows", and the rows that follow once a bind-off splits a piece at a neckline: each side is worked on its own count, so those rows are named as unread rather than checked against both sides at once.
- A parenthesized group its line leaves ambiguous, such as stitches in parentheses with no count after them: "(K2, p2) to end". Parentheses are also how a pattern writes its sizes, and one line cannot be read both ways, so the blurred cases are named as unread rather than guessed.
- Rows whose stitch count cannot be established, usually because an earlier row could not be read. Those rows are named, not skipped silently.
- Prose that changes the fabric without saying by how much. "Work as for the back, reversing shaping" is called out by name, pointing at the section it mirrors, but the mirrored rows stay unread rather than guessed.
- Anything a schematic would need beyond stitch counts: a schematic comes from graded measurements, never from the written rows.
What an unread row means
The checker follows a running stitch count. If it cannot read a row and that leaves the count unknown, later rows in that size may also need review because their starting count is unknown.
That is still a useful partial check: rows the checker did read, and any sizes it could follow, keep their own results. For the best coverage, put one instruction on each written row and state stitch counts clearly before and after shaping.
Three verdicts, and why they are not two
Warning means the checker read the row and the arithmetic looks wrong. Somebody should look at it.
Needs review means the checker could not read the construct at all. It is not a complaint about your pattern; it is a gap in ours.
Merging the two would let "we didn't check this" hide inside "we checked this", which is the one thing a verifier must never do.
Found something it should read and doesn't? Tell us with the row that broke it: reporting a gap needs a free account, so we can write back.