Answer prose round 1: 1 🔴, 6 🟡 (0 misread); ::listBreak as shipped, no E 3 anchor since none verifies, plainer reasons

docs/score-anchors.md

@@ -1,7 +1,8 @@
 # Score anchors
 
-Shipped items that anchor each level of L, E and I in `todo.md` (`~/.claude/rules/todo-items.md`).
-A level no shipped item shows has no row.
+To score an item's L, E or I in `todo.md`, compare it with each level's anchors, the shipped items
+below, and give it the level whose anchors it matches. E 3, I 4 and I 5 have no row: no shipped
+item fits them.
 
 | Level | Items |
 |---|---|
@@ -11,37 +12,34 @@ A level no shipped item shows has no row.
 | L 4 | 43, 61 |
 | E 1 | 62, 103 |
 | E 2 | 71, 121 |
-| E 3 | 61, 65 |
 | I 1 | 92, 126 |
 | I 2 | 120, 123 |
 | I 3 | 49, 114 |
 
-Why each sits at its level, with its chunk's lines changed as `score-outcomes` measures them:
+Why each sits at its level:
 
-- L 1, item 90 (6 lines in 2 files): the README's tagline and `package.json`'s description.
+- L 1, item 90 (6 lines changed in 2 files): the README's tagline and `package.json`'s description.
 - L 1, item 117 (13 lines in 1 file): one decision's wording.
-- L 2, item 102 (162 lines in 5 files): the portable task-list split, in its own component.
-- L 2, item 120 (87 lines in 4 files): one loop in the block reader, by the linear-time pattern
-  items before it set.
+- L 2, item 102 (162 lines in 5 files): the portable reader's split of a task list, in
+  `src/markdown/portable/task-list.ts`.
+- L 2, item 120 (87 lines in 4 files): one loop in the block reader, following the linear-time
+  pattern earlier items set.
 - L 3, item 114 (145 lines in 15 files): every reader and both writers.
 - L 3, item 122 (285 lines in 9 files): the inline reader, the portable writer and the image module.
-- L 4, item 43 (3192 lines in 37 files): a new reader per markdown input, across `src/markdown/`.
+- L 4, item 43 (3192 lines in 37 files): a new reader per markdown flavour, across `src/markdown/`.
 - L 4, item 61 (547 lines in 9 files): a new pattern, the writer asking the reader how a line reads
   back, across `emit/` and `parse/`.
-- E 1, item 62: a move between directories, which the types, the tests and the import boundaries
-  fail on.
-- E 1, item 103: one error code folded into another, which the types and the tests fail on.
-- E 2, item 71: the README's comparison table, each cell checked by hand against the other package.
-- E 2, item 121: a decision's text, which only a reader checks.
-- E 3, item 61: its escape walk scans for an inline HTML terminator at every opener, quadratic in
-  the text; that shipped in 0.2.0 and item 96 fixed it after.
-- E 3, item 65: inside a link's text, its reading drops an image's URL and title; that shipped in
-  0.2.0 and item 122 fixed it after.
+- E 1, item 62: a move between directories; a slip in it fails the types, the tests or the import
+  boundaries.
+- E 1, item 103: one error code folded into another; a slip in it fails the types or the tests.
+- E 2, item 71: the README's comparison table, whose cells only a person checks against each of the
+  seven packages it compares.
+- E 2, item 121: a decision's text, which only review checks.
 - I 1, item 92: a working rule in `AGENTS.md`.
 - I 1, item 126: the lint config, which only contributors run.
 - I 2, item 120: one reader's speed on one list shape.
 - I 2, item 123: link titles in one writer.
-- I 3, item 49: how every reader parts lists, and the retired `!adf:listBreak` in every consumer's
-  stored markdown.
+- I 3, item 49: a breaking change to how every reader parts lists, retiring the `::listBreak`
+  separator that consumers' stored markdown held.
 - I 3, item 114: a breaking change to code text in every reader, with a migration for stored
   markdown.