@graffiticode/l0175 0.2.0 → 0.3.1

package/dist/static/usage-guide.md

@@ -7,7 +7,16 @@ Agent-facing guide for authoring L0175 programs. Read this before composing a `c
 
 ## Overview
 
-L0175 is a content-composition language for 5th-grade English Language Arts assessment items (Smarter Balanced spec ELA · Grade 5 · Claim 1 · Reasoning & Evidence). One language serves **multiple learning targets**; **every program first declares a top-level `target`**: `c1-t4` (Target 4 — *literary* texts, RL standards, dimensions like character/theme/point-of-view) or `c1-t11` (Target 11 — *informational* texts, RI standards, dimensions like relationships-interactions/author-use-of-information/point-of-view/purpose). Choose the target from the request (literary vs. informational text and the skill assessed); the dimensions, standards, and stem catalog (`stems.md`) are then scoped to that target, and mixing targets' vocabularies is a compile error. It is **item-first**: after picking the target you compose the questions (`outcome`s) first — each with a unique `id`, a `focus` naming its correct claim, and an explicit `stem` (and `stem-b` on EBSR) instantiated from the guideline's Appropriate-Stem catalog (`stems.md`) — then author the supported `claim`s and a *superset* of distractor `claim`s, each tagging the question(s) it foils via `targets`, plus evidence `source`s. The compiler then *composes* each outcome deterministically: it takes the correct claim from `focus`, draws that question's foils ONLY from the distractors that `targets` it, uses the authored stem, and assembles a finished item in one of three task models: `ebsr` (two-part evidence-based selected response), `hot-text` (select-text), or `short-text` (constructed response). One passage + superset can yield several items, each with its own bound foil set. The compiler performs no generation and no stem synthesis — it selects, validates against the guideline, and warns when a question's pool falls short. Distractors are tagged by the SBAC error taxonomy (Part A: `misreads-detail`, `erroneous-inference`, `faulty-reasoning`; Part B: `supports-wrong-claim`, `irrelevant`), each carrying a rationale; composition picks foils for error-type coverage and couples Part B evidence to the claims it plausibly supports. **For each EBSR/Hot-Text question author at least 5 viable distractors that `targets` it (aim for 5–8, over-generating since some are filtered as near-duplicates or accidentally correct) — covering all three error types with ≥2 alternatives in at least two of them, and giving each a `plausibility` score (0–1). An item draws only 3 foils, so a richer targeted pool yields stronger items; fewer than 3 targeting a question is a hard error, fewer than 5 a warning. Likewise, for EBSR Part B author at least 5 non-supporting evidence lines (`supports-wrong-claim` + `irrelevant`) so the compiler can choose the most tempting 3 foil options. No-giveaway rule: at least one of those `supports-wrong-claim` lines must list BOTH the correct claim's id AND a distractor's id in its `supports` (a line that seems to support the right answer but actually backs a misreading) — otherwise the correct evidence line stands alone, Part B telegraphs Part A, and the compiler warns "possible A↔B giveaway." Do not make every wrong-claim line point only at distractors.**
+L0175 is a content-composition language for 5th-grade English Language Arts assessment items (Smarter Balanced spec ELA · Grade 5 · Claim 1 · Reasoning & Evidence). One language serves **multiple learning targets**; **every program first declares a top-level `target`**: `c1-t4` (Target 4 — *literary* texts, RL standards, dimensions like character/theme/point-of-view) or `c1-t11` (Target 11 — *informational* texts, RI standards, dimensions like relationships-interactions/author-use-of-information/point-of-view/purpose). Choose the target from the request (literary vs. informational text and the skill assessed); the dimensions, standards, and stem catalog (`stems.md`) are then scoped to that target, and mixing targets' vocabularies is a compile error. It is **item-first**: after picking the target you compose the questions (`outcome`s) first — each with a unique `id`, a `focus` naming its correct claim, and an explicit `stem` (and `stem-b` on EBSR) instantiated from the guideline's Appropriate-Stem catalog (`stems.md`) — then author the supported `claim`s and a *superset* of distractor `claim`s, each tagging the question(s) it foils via `targets`, plus evidence `source`s. The compiler then *composes* each outcome deterministically: it takes the correct claim from `focus`, draws that question's foils ONLY from the distractors that `targets` it, uses the authored stem, and assembles a finished item in the task model named by its item type — for Reasoning & Evidence (T4/T11): `ebsr` (two-part evidence-based selected response), `hot-text` (select-text), or `short-text` (constructed response); other targets also offer `multiple-choice`/`multi-select` (the allowed set and the per-target task-model numbering are in *Vocabulary Cues*). One passage + superset can yield several items, each with its own bound foil set. The compiler performs no generation and no stem synthesis — it selects, validates against the guideline, and warns when a question's pool falls short. Distractors are tagged by the SBAC error taxonomy (Part A: `misreads-detail`, `erroneous-inference`, `faulty-reasoning`; Part B: `supports-wrong-claim`, `irrelevant`), each carrying a rationale; composition picks foils for error-type coverage and couples Part B evidence to the claims it plausibly supports. **For each EBSR/Hot-Text question author at least 5 viable distractors that `targets` it (aim for 5–8, over-generating since some are filtered as near-duplicates or accidentally correct) — covering all three error types with ≥2 alternatives in at least two of them, and giving each a `plausibility` score (0–1). An item draws only 3 foils, so a richer targeted pool yields stronger items; fewer than 3 targeting a question is a hard error, fewer than 5 a warning. Likewise, for EBSR Part B author at least 5 non-supporting evidence lines (`supports-wrong-claim` + `irrelevant`) so the compiler can choose the most tempting 3 foil options. No-giveaway rule: at least one of those `supports-wrong-claim` lines must list BOTH the correct claim's id AND a distractor's id in its `supports` (a line that seems to support the right answer but actually backs a misreading) — otherwise the correct evidence line stands alone, Part B telegraphs Part A, and the compiler warns "possible A↔B giveaway." Do not make every wrong-claim line point only at distractors.**
+
+**Name the item type; task-model NUMBERS are per-target and collide.** Phrase a request by item
+type (`ebsr` / `hot-text` / `short-text` / `multiple-choice` / `multi-select`) — that is unambiguous.
+Task-model *numbers* (TM1…TM5) are numbered **per target**, so the same number means different things
+in different targets: `tm3` is **short-text** in `c1-t4`/`c1-t11`, **`ebsr`** in `c1-t9`, and
+**`hot-text`** in `c1-t1`/`c1-t8`/`c1-t10`. A bare "task model 3" is therefore meaningless without
+its target — never assume the Reasoning & Evidence (T4/T11) numbering carries over to the others. When a
+request names a task model by number, resolve it against the **program's target** (see the per-target
+task-model table in *Vocabulary Cues*) and state the item type, e.g. for `c1-t9` "task model 3 → EBSR".
 
 **Targets are different skills — pick the one the request assesses.** Beyond the two Reasoning &
 Evidence targets above (`c1-t4` literary, `c1-t11` informational — infer/conclude and justify with
@@ -26,6 +35,16 @@ correct selection). Two new item types — `multiple-choice` (4 options, one cor
 `multi-select` (5–6 options, an exact correct set) — are single-part: author the `stem` and a
 `focus`, no Part B.
 
+**`c1-t2` — Central Ideas over LITERARY texts** is T9's twin on a story, with the same five item
+types, the same significance taxonomy, and standards `rl-1`+`rl-2`. Choose it over `c1-t9` by the
+passage: story/poem → `c1-t2`, article → `c1-t9`. Two rules are T2's alone. **Theme leads**: its
+dimensions are `theme`, `central-idea`, `key-detail`, `summary`, and for a story you normally want
+`theme` with the correct answer phrased as a *lesson or message* ("Keeping at something can matter
+more than winning"), not a plot event — reserve `central-idea` for a prompt that says "main idea".
+**Summary is scoped**: the guideline forbids asking students to summarize the *entire* text, so a
+T2 `summary` stem must name a section or key event ("Summarize what happens after…"). A plain
+"summarize the passage" belongs to T9, not T2.
+
 L0175 also composes **`c1-t8` — Key Details** (informational): the inference/conclusion is **GIVEN
 in the stem** and the student selects the supporting **evidence** (the answer is *evidence*, not a
 chosen statement). Use `c1-t8` when the request **states an idea and asks which detail/sentence
@@ -35,6 +54,13 @@ supports it**; dimension `supporting-evidence`, standards `ri-1`+`ri-7`, DOK 1
 options: `directly-supports` = correct evidence (with a `quote`), `supports-wrong-claim`/`irrelevant`
 = distractor evidence. No distractor claims.
 
+**`c1-t1` — Key Details over LITERARY texts** is the same model as `c1-t8` on a story rather than an
+article: same dimension `supporting-evidence`, same item types, same DOK 1–2, same authoring recipe.
+The one difference is the standard — the guideline names **`rl-1` and no companion**, so a composed
+item's `standards` is exactly `["rl-1"]`; omit `standard` on the outcome. Choose between them by the
+passage: story/poem/narrative → `c1-t1`, article/report → `c1-t8`. T1's stems offer `line` as a
+selectable unit and say "[author/**narrator**]"; its Multi-Select is exactly two correct.
+
 L0175 also composes **`c1-t10` — Word Meanings** (informational): the question asks for the
 **meaning of a targeted word/phrase in context**, so the answer choices are **meanings**, authored
 as a dedicated `word`/`meaning` structure (not claims). Use `c1-t10` when the request asks **what a
@@ -71,7 +97,7 @@ Author the passage **and** every claim, option, and rationale to that grade (fig
 
 Say this to get that:
 
-- **Target** — `target c1-t4` (literary, Reasoning & Evidence), `target c1-t11` (informational, Reasoning & Evidence), `target c1-t9` (informational, Central Ideas), `target c1-t8` (informational, Key Details — given-inference → evidence), or `target c1-t10` (informational, Word Meanings — meaning of a targeted word); required, first top-level form. Selects the dimensions, standards, distractor taxonomy, DOK, item types, and stem catalog.
+- **Target** — `target c1-t4` (literary, Reasoning & Evidence), `target c1-t11` (informational, Reasoning & Evidence), `target c1-t2` (**literary**, Central Ideas), `target c1-t9` (informational, Central Ideas), `target c1-t1` (**literary**, Key Details — given-inference → evidence), `target c1-t8` (informational, Key Details — same model), or `target c1-t10` (informational, Word Meanings — meaning of a targeted word); required, first top-level form. Selects the dimensions, standards, distractor taxonomy, DOK, item types, and stem catalog.
 - **Grade** — optional top-level `grade <n>` (e.g. `grade 5`). Defaults to the target/guideline's grade; author one only to override when the user asks for a different grade. Sets the reading-level target the compiler checks the passage against.
 - **Passage** — `passage "Title"` sets the heading; `type literary` (or `informational`, matching the target); `lines [ "..." "..." ]` are the passage **paragraphs** — one entry per paragraph, auto-numbered from 1. Always split by paragraph, including for Hot Text: the compiler segments each paragraph into sentences and exposes each sentence as a Part B selection, so the passage keeps its paragraph format. For EBSR Part B, evidence sources carry a `quote` with the exact supporting sentence so options stay sentence-tight while `line` points at the paragraph.
 - **Outcome** — the question, composed first. `outcome id "q1" type ebsr dimension character subject "Mara" focus "c1" stem "Which of these inferences about Mara is supported by the passage?" stem-b "Which sentence(s) from the passage best support your answer in Part A?" {}`. `id` is the handle distractors target; `focus` names the correct claim; `stem` (and `stem-b` on EBSR) come from `stems.md`. Vary `type` (`ebsr` / `hot-text` / `short-text`) for different task models. **Keep the Part A stem a neutral question** — fill the template's slot with the subject/skill only, and do **not** echo the correct option's wording (if the stem restates the answer, it gives it away). Paraphrase so stem and key share only the subject; the compiler warns when the stem reuses most of the correct option's words. Applies to every Hot Text and EBSR question, across all targets. **For Hot Text, Part A asks for the best STATEMENT** ("Click on the statement that best provides an inference about [X]…") whose options are inference claims — it is NOT a "click the sentences" instruction (that's Part B, fixed by the compiler). If a request says "select the sentences that show [X]", translate it: Part A is a statement prompt about [X], and the sentences that show [X] are `directly-supports` evidence (with exact `quote`s) that become Part B's answer. The compiler warns when a Hot Text Part A stem mentions sentences.
@@ -79,12 +105,26 @@ Say this to get that:
 - **Evidence source** — a passage line tagged by its support role. `source id "e1" line 1 status directly-supports supports ["c1"] {}`. Statuses: `directly-supports`, `supports-wrong-claim`, `irrelevant`. An optional `rationale` explains a foil. For **Hot Text**, give each `directly-supports` source a `quote` naming the exact supporting **sentence** so that sentence is marked correct in Part B; a `directly-supports` source with no `quote` marks every sentence of its `line` correct. Part B asks for an **exact number** of sentences (one less than the valid count, floored at 1, capped at 3) and **any selection of that many from the valid set is correct** (a superset) — so **author ≥3 directly-supporting sentences** so the asked count is ≥2 and students aren't forced to find every one.
 - **Dimensions (c1-t4)** — `character`, `setting`, `event`, `point-of-view`, `theme`, `topic`, `narrators-feelings`, `character-relationship`.
 - **Dimensions (c1-t11)** — `relationships-interactions`, `author-use-of-information`, `point-of-view`, `purpose`, `authors-opinion`.
+- **Dimensions (c1-t2)** — `theme` (prefer it for a story: the lesson/message), `central-idea`, `key-detail` (covers key events), `summary` (scoped to a section or key event, never the whole text).
 - **Dimensions (c1-t9)** — `central-idea`, `key-detail`, `summary`.
-- **Dimensions (c1-t8)** — `supporting-evidence` (the answer is evidence; the inference is given in the stem).
+- **Dimensions (c1-t1 / c1-t8)** — `supporting-evidence` (the answer is evidence; the inference is given in the stem). `c1-t1` is the literary target, `c1-t8` the informational one.
 - **Dimensions (c1-t10)** — `word-meaning` (the answer is a meaning; authored via `word`/`meaning`).
 - **Word / meaning (c1-t10)** — `words [ word id "w1" text "aqueduct" line 1 quote "…" meanings [ meaning id "m1" status correct text "a water channel" {} meaning id "m2" status distractor error-type other-meaning text "a boat" rationale "…" {} ] {} ]`; the outcome's `focus` names the word.
 - **Item types** — `ebsr` (two-part: statement → evidence), `hot-text` (R&E/Central-Ideas: click the supporting/main-idea sentences; Key Details: click the evidence sentences; Word Meanings: click the word matching a definition), `short-text` (constructed response), `multiple-choice` (one correct, single-part), `multi-select` (exact correct set, single-part). MC/Multi-Select have no Part B. The allowed set is per-target (the compiler rejects others).
-- **Central Ideas (c1-t9) distractor `error-type`** — `too-narrow`, `too-broad`, `misreads-detail`, `insignificant` (true-but-not-central); R&E targets use `misreads-detail`/`erroneous-inference`/`faulty-reasoning`.
+- **Task models are per-target — resolve a number against the target.** Task-model NUMBERS (TM1…TM5) are numbered per target and collide, so name the item type rather than a bare number; if a request says "task model N", map N to the item type for the program's target using the table below (e.g. `c1-t9` TM3 → `ebsr`, TM4 → `hot-text`). **Specify the task model on the outcome as `task-model tm3` whenever it is known** — the compiler resolves it against the target's table and **hard-errors if it disagrees with `type`** (or supplies `type` when omitted), so the intended task model is checked, not guessed. It stays optional, but the compiler emits a (non-blocking) warning when an outcome omits it, so include it whenever you know it.
+
+<!-- GENERATED:task-models START (from targets.json — regenerated by tools/build-static.js; do not edit by hand) -->
+| Target | tm1 | tm2 | tm3 | tm4 | tm5 |
+|--------|-----|-----|-----|-----|-----|
+| `c1-t4` — Grade 5 · Claim 1 · Target 4 (Reasoning & Evidence) | ebsr | hot-text | short-text | — | — |
+| `c1-t11` — Grade 5 · Claim 1 · Target 11 (Reasoning & Evidence) | ebsr | hot-text | short-text | — | — |
+| `c1-t2` — Grade 5 · Claim 1 · Target 2 (Central Ideas) | multiple-choice | multi-select | ebsr | hot-text | short-text |
+| `c1-t9` — Grade 5 · Claim 1 · Target 9 (Central Ideas) | multiple-choice | multi-select | ebsr | hot-text | short-text |
+| `c1-t1` — Grade 5 · Claim 1 · Target 1 (Key Details) | multiple-choice | multi-select | hot-text | — | — |
+| `c1-t8` — Grade 5 · Claim 1 · Target 8 (Key Details) | multiple-choice | multi-select | hot-text | — | — |
+| `c1-t10` — Grade 5 · Claim 1 · Target 10 (Word Meanings) | multiple-choice | multi-select | hot-text | — | — |
+<!-- GENERATED:task-models END -->
+- **Central Ideas (c1-t2 / c1-t9) distractor `error-type`** — `too-narrow`, `too-broad`, `misreads-detail`, `insignificant` (true-but-not-central); R&E targets use `misreads-detail`/`erroneous-inference`/`faulty-reasoning`. Keep an `insignificant` foil plausible — the guideline warns against details so trivial the option is implausible.
 - **Program terminator** — top-level forms chain with no `{}` between them; the program ends with a single `{}..`.
 
 ## Example Prompts
@@ -95,16 +135,19 @@ Say this to get that:
 - *"Make an EBSR item about how the author uses evidence to support a point in an informational article about bridges, standard ri-8."* → `target c1-t11`, `dimension author-use-of-information`
 - *"Write an item about the relationships between the events in a history passage."* → `target c1-t11`, `dimension relationships-interactions`
 - *"Which sentence best states the main idea of this article? (multiple choice)"* → `target c1-t9`, `type multiple-choice`, `dimension central-idea`
+- *"What is the theme of this story? Make it a two-part item with the evidence in Part B."* → `target c1-t2`, `type ebsr` (tm3), `dimension theme`; the correct answer is a lesson, not a plot event
+- *"Have students summarize what happens after the storm, using details from the story."* → `target c1-t2`, `type short-text` (tm5), `dimension summary` — scoped to a key event, which is the only kind of summary T2 allows
 - *"Make a multi-select asking for the two sentences that belong in a summary."* → `target c1-t9`, `type multi-select`, `dimension summary`, `focus` is a list of the two correct claims
 - *"The article says aqueducts brought water to distant cities — which detail best supports that? (multiple choice)"* → `target c1-t8`, `type multiple-choice`, `dimension supporting-evidence`; the inference goes in the `stem`, the options are `source`s
 - *"Make a click-the-sentence item where students select the evidence for a stated conclusion."* → `target c1-t8`, `type hot-text` (single-part), `dimension supporting-evidence`
+- *"In this story, the reader can conclude the boy regrets what he said — which line best supports that?"* → `target c1-t1` (the literary Key Details target), `type multiple-choice`, `dimension supporting-evidence`; standards are `rl-1` alone
 - *"What does the word 'aqueduct' mean as used in this article? (multiple choice)"* → `target c1-t10`, `type multiple-choice`, `dimension word-meaning`; a `word` with candidate `meaning`s, `focus` names the word
 - *"Read the paragraph below. Click the word that means 'a channel that carries water'."* → `target c1-t10`, `type hot-text` (click-the-word); author the candidate `word`s (correct one is `focus` with its paragraph's `line`, plus a few distractor candidate words from the same paragraph); definition in the stem, passage NOT in the stem
 - *"Keep the passage at a 4th-grade reading level."* → top-level `grade 4`; author shorter sentences and simpler vocabulary (the default is the guideline's Grade 5)
 
 ## Out of Scope
 
-- **Other targets / grades / claims** — L0175 covers G5 · Claim 1 · Reasoning & Evidence, targets T4 (literary) and T11 (informational). Other targets belong in their own dialects.
+- **Other targets / grades / claims** — L0175 covers G5 · Claim 1, targets T4 (Reasoning & Evidence, literary), T11 (Reasoning & Evidence, informational), T9 (Central Ideas), T8 (Key Details), and T10 (Word Meanings). Other claims, grades, or Claim-1 targets belong in their own dialects.
 - **Dual-text stimuli** — a single passage only in this version.
 - **Compile-time generation** — the compiler selects and validates authored content; it does not invent claims, distractors, or evidence.
 - **Auto-scoring** — short-text responses are hand-scored against the rubric; the compiler emits the rubric only.

package/dist/targets.d.ts

@@ -0,0 +1,27 @@
+export declare const TARGETS_REVISED = "2026-08-28";
+export declare const STANDARD_FAMILIES: {
+    readonly RL_G5: readonly ["rl-1", "rl-2", "rl-3", "rl-4", "rl-5", "rl-6", "rl-7", "rl-9"];
+    readonly RI_G5: readonly ["ri-1", "ri-2", "ri-3", "ri-4", "ri-5", "ri-6", "ri-7", "ri-8", "ri-9"];
+    readonly L_G5: readonly ["l-4", "l-4a", "l-4b", "l-4c", "l-5", "l-5a", "l-5b", "l-5c"];
+};
+export type StandardFamily = keyof typeof STANDARD_FAMILIES;
+export type TargetData = {
+    id: string;
+    label: string;
+    grade: number;
+    skill: "reasoning-evidence" | "central-ideas" | "key-details" | "word-meanings";
+    textType: "literary" | "informational";
+    baseStandard: string;
+    defaultDok: string;
+    answerKind: "statement" | "evidence" | "meaning";
+    singlePartHotText: boolean;
+    taskModels: Record<string, string>;
+    standards: StandardFamily[];
+    dimensions: string[];
+    errorTypes: string[];
+    dimStandard: Record<string, string>;
+};
+export declare function targetForSkill(skill: string, textType?: string): TargetData | undefined;
+export declare function taskModelNumber(target: string, itemType: string): string | undefined;
+export declare const TARGETS_DATA: Record<string, TargetData>;
+//# sourceMappingURL=targets.d.ts.map

package/dist/targets.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"targets.d.ts","sourceRoot":"","sources":["../src/targets.ts"],"names":[],"mappings":"AAmBA,eAAO,MAAM,eAAe,eAAe,CAAC;AAM5C,eAAO,MAAM,iBAAiB;;;;CAIpB,CAAC;AAEX,MAAM,MAAM,cAAc,GAAG,MAAM,OAAO,iBAAiB,CAAC;AAI5D,MAAM,MAAM,UAAU,GAAG;IACvB,EAAE,EAAE,MAAM,CAAC;IACX,KAAK,EAAE,MAAM,CAAC;IACd,KAAK,EAAE,MAAM,CAAC;IAKd,KAAK,EAAE,oBAAoB,GAAG,eAAe,GAAG,aAAa,GAAG,eAAe,CAAC;IAChF,QAAQ,EAAE,UAAU,GAAG,eAAe,CAAC;IACvC,YAAY,EAAE,MAAM,CAAC;IACrB,UAAU,EAAE,MAAM,CAAC;IAInB,UAAU,EAAE,WAAW,GAAG,UAAU,GAAG,SAAS,CAAC;IAGjD,iBAAiB,EAAE,OAAO,CAAC;IAK3B,UAAU,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IACnC,SAAS,EAAE,cAAc,EAAE,CAAC;IAC5B,UAAU,EAAE,MAAM,EAAE,CAAC;IACrB,UAAU,EAAE,MAAM,EAAE,CAAC;IACrB,WAAW,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;CACrC,CAAC;AAMF,wBAAgB,cAAc,CAAC,KAAK,EAAE,MAAM,EAAE,QAAQ,CAAC,EAAE,MAAM,GAAG,UAAU,GAAG,SAAS,CAKvF;AAMD,wBAAgB,eAAe,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,GAAG,MAAM,GAAG,SAAS,CAKpF;AAoBD,eAAO,MAAM,YAAY,EAAE,MAAM,CAAC,MAAM,EAAE,UAAU,CAqLnD,CAAC"}

package/dist/targets.js

@@ -0,0 +1,250 @@
+// SPDX-License-Identifier: MIT
+/* Copyright (c) 2026, ARTCOMPILER INC */
+//
+// SINGLE SOURCE OF TRUTH for L0175's per-target structure.
+//
+// Everything the compiler enforces about a learning target lives here — and ONLY here.
+// `compiler.ts` builds its runtime `TARGETS` profiles from this data; `tools/build-static.js`
+// emits a served `targets.json` from it and GENERATES the per-target tables in the spec
+// `.md` files (instructions.md, stems.md) from it. The `.md` files are the generator-facing
+// flip side of this data: they carry MORE (narrative, stems, examples) but must EMPHASIZE
+// everything here — nothing structural the compiler checks may be absent from what the
+// generator reads. To add or change a target fact, edit this file; the compiler and the docs
+// follow.
+//
+// This is plain data (no logic) so it round-trips to JSON. It is a typed `.ts` module rather
+// than a raw `.json` import because core compiles to ESM under NodeNext, where JSON imports
+// need version-fragile import attributes; build-static serializes it to `dist/static/targets.json`
+// for non-TS consumers (the console RAG pipeline) and humans.
+export const TARGETS_REVISED = "2026-08-28";
+// CCSS Grade-5 reading-standard families. A target's `standards` is the full strand for its
+// text type, so any plausible CCSS code for that text type validates (the `dimStandard` map
+// still picks the right companion by default). RL.8 is "not applicable to literature";
+// RL.10/RI.10 are range-of-reading bands, not discrete item standards.
+export const STANDARD_FAMILIES = {
+    RL_G5: ["rl-1", "rl-2", "rl-3", "rl-4", "rl-5", "rl-6", "rl-7", "rl-9"],
+    RI_G5: ["ri-1", "ri-2", "ri-3", "ri-4", "ri-5", "ri-6", "ri-7", "ri-8", "ri-9"],
+    L_G5: ["l-4", "l-4a", "l-4b", "l-4c", "l-5", "l-5a", "l-5b", "l-5c"], // vocabulary (T10)
+};
+// Route a (skill, textType) pair to its target — the skill × text-type grid the docs present to
+// the generator. `textType` may be omitted when a request names a skill but no genre; the literary
+// target wins then, per the spec's "when the text type is ambiguous, prefer the literary target",
+// falling back to the skill's only form (Word Meanings has no literary target).
+export function targetForSkill(skill, textType) {
+    const all = Object.values(TARGETS_DATA).filter((d) => d.skill === skill);
+    if (!all.length)
+        return undefined;
+    if (textType)
+        return all.find((d) => d.textType === textType) ?? all[0];
+    return all.find((d) => d.textType === "literary") ?? all[0];
+}
+// Reverse lookup: the task-model NUMBER for a (target, item type) — e.g. ("c1-t9", "ebsr") → "3",
+// ("c1-t4", "short-text") → "3". Item-type → task-model is per-target (the numbers collide), so the
+// target is required. Returns undefined if the target doesn't offer that item type. Used by the RAG
+// signature so a number-phrased query ("task model 3") can match a target's correct exemplar.
+export function taskModelNumber(target, itemType) {
+    const tms = TARGETS_DATA[target]?.taskModels;
+    if (!tms)
+        return undefined;
+    const key = Object.keys(tms).find((k) => tms[k] === itemType);
+    return key ? key.replace(/^tm/, "") : undefined;
+}
+// Distractor error taxonomies reused across targets. Reasoning & Evidence (T4/T11) classifies
+// foils by reasoning failure; Central Ideas (T9) by SIGNIFICANCE (a true statement that just
+// isn't the central idea); Word Meanings (T10) by meaning error.
+const RE_ERROR_TYPES = ["misreads-detail", "erroneous-inference", "faulty-reasoning"];
+// Shared by BOTH Central Ideas targets (T9 informational, T2 literary): their guidelines describe
+// the same two distractor failures — misinterpreting a detail, and misjudging the SIGNIFICANCE of
+// ideas and details — which this taxonomy splits into a misread plus three significance errors.
+const CENTRAL_IDEA_ERROR_TYPES = ["too-narrow", "too-broad", "misreads-detail", "insignificant"];
+// R&E (T4/T11) and the single-select families share task-model layouts.
+const RE_TASK_MODELS = { tm1: "ebsr", tm2: "hot-text", tm3: "short-text" };
+const SELECT_TASK_MODELS = { tm1: "multiple-choice", tm2: "multi-select", tm3: "hot-text" };
+// Central Ideas (T9/T2) is the only family that offers all five item types, so its numbering
+// reaches tm5. NOTE tm3 is EBSR here but hot-text in the SELECT_TASK_MODELS targets.
+const CENTRAL_IDEA_TASK_MODELS = {
+    tm1: "multiple-choice", tm2: "multi-select", tm3: "ebsr", tm4: "hot-text", tm5: "short-text",
+};
+export const TARGETS_DATA = {
+    // Claim 1 · Target 4 — Reasoning & Evidence, literary texts (RL standards). The original L0175.
+    "c1-t4": {
+        id: "c1-t4",
+        label: "Grade 5 · Claim 1 · Target 4 (Reasoning & Evidence)",
+        grade: 5,
+        skill: "reasoning-evidence",
+        textType: "literary",
+        baseStandard: "rl-1",
+        defaultDok: "r-dok3",
+        answerKind: "statement",
+        singlePartHotText: false,
+        taskModels: RE_TASK_MODELS,
+        standards: ["RL_G5"],
+        dimensions: [
+            "character", "setting", "event", "point-of-view",
+            "theme", "topic", "narrators-feelings", "character-relationship",
+        ],
+        errorTypes: RE_ERROR_TYPES,
+        dimStandard: {
+            "character": "rl-3", "character-relationship": "rl-3", "setting": "rl-3", "event": "rl-3",
+            "point-of-view": "rl-6", "narrators-feelings": "rl-6",
+            // theme/topic = determine-the-theme/summarize → RL.2 (the CCSS theme standard), not the
+            // cross-text-comparison RL.9.
+            "theme": "rl-2", "topic": "rl-2",
+        },
+    },
+    // Claim 1 · Target 11 — Reasoning & Evidence, informational texts (RI standards).
+    "c1-t11": {
+        id: "c1-t11",
+        label: "Grade 5 · Claim 1 · Target 11 (Reasoning & Evidence)",
+        grade: 5,
+        skill: "reasoning-evidence",
+        textType: "informational",
+        baseStandard: "ri-1",
+        defaultDok: "r-dok3",
+        answerKind: "statement",
+        singlePartHotText: false,
+        taskModels: RE_TASK_MODELS,
+        standards: ["RI_G5"],
+        dimensions: [
+            "relationships-interactions", "author-use-of-information",
+            "point-of-view", "purpose", "authors-opinion",
+        ],
+        errorTypes: RE_ERROR_TYPES,
+        dimStandard: {
+            "relationships-interactions": "ri-3",
+            "author-use-of-information": "ri-8",
+            "point-of-view": "ri-6",
+            "purpose": "ri-8",
+            "authors-opinion": "ri-8",
+        },
+    },
+    // Claim 1 · Target 2 — Central Ideas, LITERARY texts (RL-1 + RL-2). The literary twin of T9, and
+    // the only other target offering all five item types. Its own emphases, from the guideline:
+    //   - THEME leads. RL-2 is "determine a theme of a story, drama, or poem from details in the
+    //     text … summarize the text", so `theme` is a first-class dimension here, not a T4 borrowing.
+    //   - SUMMARY IS SCOPED. "Items will not ask students to summarize the entire text; students will
+    //     summarize a key event(s) or idea(s)." T9 has no such limit — this is the sharpest T2/T9
+    //     difference and the docs must carry it.
+    //   - `key-detail` covers the guideline's "key event(s) or idea(s)" — in a story the key details
+    //     that build a theme are usually events, so they share one dimension rather than splitting a
+    //     single skill across two tags the compiler treats identically.
+    // Every dimension answers to RL-2 (the theme/summarize standard); RL-1 is added as the base, so a
+    // composed item's standards is ["rl-1", "rl-2"] — exactly the guideline's pair.
+    "c1-t2": {
+        id: "c1-t2",
+        label: "Grade 5 · Claim 1 · Target 2 (Central Ideas)",
+        grade: 5,
+        skill: "central-ideas",
+        textType: "literary",
+        baseStandard: "rl-1",
+        defaultDok: "r-dok2", // tm5 short-text bumps to r-dok3 (dokFor), per the guideline
+        answerKind: "statement",
+        singlePartHotText: true, // T2 Hot Text (tm4): click the sentence(s) that show the theme
+        taskModels: CENTRAL_IDEA_TASK_MODELS,
+        standards: ["RL_G5"],
+        dimensions: ["theme", "central-idea", "key-detail", "summary"],
+        errorTypes: CENTRAL_IDEA_ERROR_TYPES,
+        dimStandard: {
+            "theme": "rl-2",
+            "central-idea": "rl-2",
+            "key-detail": "rl-2",
+            "summary": "rl-2",
+        },
+    },
+    // Claim 1 · Target 9 — Central Ideas, informational texts (RI-1 + RI-2). A DIFFERENT skill from
+    // Reasoning & Evidence: whole-text synthesis and significance (the main idea, the key details that
+    // build it, and summary), DOK 2 (3 only for the written summary). Distractors are a SIGNIFICANCE
+    // taxonomy — usually true statements that just aren't the central idea. The only non-R&E target
+    // that allows EBSR and short-text, so its task-model numbering reaches tm5.
+    "c1-t9": {
+        id: "c1-t9",
+        label: "Grade 5 · Claim 1 · Target 9 (Central Ideas)",
+        grade: 5,
+        skill: "central-ideas",
+        textType: "informational",
+        baseStandard: "ri-1",
+        defaultDok: "r-dok2",
+        answerKind: "statement",
+        singlePartHotText: true, // T9 Hot Text (tm4): click the sentence(s) that show the main idea
+        taskModels: CENTRAL_IDEA_TASK_MODELS,
+        standards: ["RI_G5"],
+        dimensions: ["central-idea", "key-detail", "summary"],
+        errorTypes: CENTRAL_IDEA_ERROR_TYPES,
+        dimStandard: {
+            "central-idea": "ri-2",
+            "key-detail": "ri-2",
+            "summary": "ri-2",
+        },
+    },
+    // Claim 1 · Target 1 — Key Details, LITERARY texts (RL-1 only). The literary mirror of T8: the
+    // inference/conclusion is GIVEN in the stem and the student selects the supporting EVIDENCE
+    // (answerKind "evidence"), same three task models. Unlike T8 the guideline lists NO companion
+    // standard — RL-1 (quote accurately / draw inferences) is the whole of it, so `dimStandard` maps
+    // back to the base and an item's `standards` is just ["rl-1"]. Its literary stems say
+    // "[author/narrator]" and add "line" as a selectable unit; Multi-Select is exactly TWO correct
+    // (T8 allows two or three). DOK 1–2 (default 2).
+    "c1-t1": {
+        id: "c1-t1",
+        label: "Grade 5 · Claim 1 · Target 1 (Key Details)",
+        grade: 5,
+        skill: "key-details",
+        textType: "literary",
+        baseStandard: "rl-1",
+        defaultDok: "r-dok2",
+        answerKind: "evidence",
+        singlePartHotText: true,
+        taskModels: SELECT_TASK_MODELS,
+        standards: ["RL_G5"],
+        dimensions: ["supporting-evidence"],
+        errorTypes: [], // like T8, wrong answers are non-supporting sources, not distractor claims
+        dimStandard: {
+            "supporting-evidence": "rl-1", // no companion in the guideline — RL-1 stands alone
+        },
+    },
+    // Claim 1 · Target 8 — Key Details, informational texts (RI-1 + RI-7). A DIFFERENT model: the
+    // inference/conclusion is GIVEN in the stem and the student selects the supporting EVIDENCE
+    // (answerKind "evidence"). Options are passage sources, not claims; no statement Part A, no
+    // EBSR/short-text. DOK 1–2 (default 2).
+    "c1-t8": {
+        id: "c1-t8",
+        label: "Grade 5 · Claim 1 · Target 8 (Key Details)",
+        grade: 5,
+        skill: "key-details",
+        textType: "informational",
+        baseStandard: "ri-1",
+        defaultDok: "r-dok2",
+        answerKind: "evidence",
+        singlePartHotText: true,
+        taskModels: SELECT_TASK_MODELS,
+        standards: ["RI_G5"],
+        dimensions: ["supporting-evidence"],
+        errorTypes: [], // T8 wrong answers are non-supporting sources, not distractor claims
+        dimStandard: {
+            "supporting-evidence": "ri-7",
+        },
+    },
+    // Claim 1 · Target 10 — Word Meanings, informational texts (RI-4 + the L-4 family). The most
+    // different model: the question asks for the MEANING of a targeted word/phrase in context, so the
+    // options are candidate MEANINGS (answerKind "meaning"), authored as `word`/`meaning`, not claims.
+    // DOK 1–2. The strategy (context / roots & affixes / word relationships / reference) is expressed
+    // via the authored standard (l-4a / l-4b / l-5c / l-4c).
+    "c1-t10": {
+        id: "c1-t10",
+        label: "Grade 5 · Claim 1 · Target 10 (Word Meanings)",
+        grade: 5,
+        skill: "word-meanings",
+        textType: "informational",
+        baseStandard: "ri-4",
+        defaultDok: "r-dok2",
+        answerKind: "meaning",
+        singlePartHotText: false, // T10 Hot Text is word-level (composeWordMeaning), not sentence-level
+        taskModels: SELECT_TASK_MODELS,
+        standards: ["RI_G5", "L_G5"],
+        dimensions: ["word-meaning"],
+        errorTypes: ["other-meaning", "misinterprets", "wrong-context"],
+        dimStandard: {
+            "word-meaning": "l-4",
+        },
+    },
+};
+//# sourceMappingURL=targets.js.map

package/dist/targets.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"targets.js","sourceRoot":"","sources":["../src/targets.ts"],"names":[],"mappings":"AAAA,+BAA+B;AAC/B,yCAAyC;AACzC,EAAE;AACF,2DAA2D;AAC3D,EAAE;AACF,uFAAuF;AACvF,8FAA8F;AAC9F,wFAAwF;AACxF,4FAA4F;AAC5F,0FAA0F;AAC1F,uFAAuF;AACvF,6FAA6F;AAC7F,UAAU;AACV,EAAE;AACF,6FAA6F;AAC7F,4FAA4F;AAC5F,mGAAmG;AACnG,8DAA8D;AAE9D,MAAM,CAAC,MAAM,eAAe,GAAG,YAAY,CAAC;AAE5C,4FAA4F;AAC5F,4FAA4F;AAC5F,uFAAuF;AACvF,uEAAuE;AACvE,MAAM,CAAC,MAAM,iBAAiB,GAAG;IAC/B,KAAK,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC;IACvE,KAAK,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC;IAC/E,IAAI,EAAE,CAAC,KAAK,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,mBAAmB;CACjF,CAAC;AAoCX,gGAAgG;AAChG,mGAAmG;AACnG,kGAAkG;AAClG,gFAAgF;AAChF,MAAM,UAAU,cAAc,CAAC,KAAa,EAAE,QAAiB;IAC7D,MAAM,GAAG,GAAG,MAAM,CAAC,MAAM,CAAC,YAAY,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,KAAK,KAAK,KAAK,CAAC,CAAC;IACzE,IAAI,CAAC,GAAG,CAAC,MAAM;QAAE,OAAO,SAAS,CAAC;IAClC,IAAI,QAAQ;QAAE,OAAO,GAAG,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,QAAQ,KAAK,QAAQ,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC,CAAC;IACxE,OAAO,GAAG,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,QAAQ,KAAK,UAAU,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC,CAAC;AAC9D,CAAC;AAED,kGAAkG;AAClG,oGAAoG;AACpG,oGAAoG;AACpG,8FAA8F;AAC9F,MAAM,UAAU,eAAe,CAAC,MAAc,EAAE,QAAgB;IAC9D,MAAM,GAAG,GAAG,YAAY,CAAC,MAAM,CAAC,EAAE,UAAU,CAAC;IAC7C,IAAI,CAAC,GAAG;QAAE,OAAO,SAAS,CAAC;IAC3B,MAAM,GAAG,GAAG,MAAM,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,KAAK,QAAQ,CAAC,CAAC;IAC9D,OAAO,GAAG,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;AAClD,CAAC;AAED,8FAA8F;AAC9F,6FAA6F;AAC7F,iEAAiE;AACjE,MAAM,cAAc,GAAG,CAAC,iBAAiB,EAAE,qBAAqB,EAAE,kBAAkB,CAAC,CAAC;AACtF,kGAAkG;AAClG,kGAAkG;AAClG,gGAAgG;AAChG,MAAM,wBAAwB,GAAG,CAAC,YAAY,EAAE,WAAW,EAAE,iBAAiB,EAAE,eAAe,CAAC,CAAC;AAEjG,wEAAwE;AACxE,MAAM,cAAc,GAAG,EAAE,GAAG,EAAE,MAAM,EAAE,GAAG,EAAE,UAAU,EAAE,GAAG,EAAE,YAAY,EAAE,CAAC;AAC3E,MAAM,kBAAkB,GAAG,EAAE,GAAG,EAAE,iBAAiB,EAAE,GAAG,EAAE,cAAc,EAAE,GAAG,EAAE,UAAU,EAAE,CAAC;AAC5F,6FAA6F;AAC7F,qFAAqF;AACrF,MAAM,wBAAwB,GAAG;IAC/B,GAAG,EAAE,iBAAiB,EAAE,GAAG,EAAE,cAAc,EAAE,GAAG,EAAE,MAAM,EAAE,GAAG,EAAE,UAAU,EAAE,GAAG,EAAE,YAAY;CAC7F,CAAC;AAEF,MAAM,CAAC,MAAM,YAAY,GAA+B;IACtD,gGAAgG;IAChG,OAAO,EAAE;QACP,EAAE,EAAE,OAAO;QACX,KAAK,EAAE,qDAAqD;QAC5D,KAAK,EAAE,CAAC;QACR,KAAK,EAAE,oBAAoB;QAC3B,QAAQ,EAAE,UAAU;QACpB,YAAY,EAAE,MAAM;QACpB,UAAU,EAAE,QAAQ;QACpB,UAAU,EAAE,WAAW;QACvB,iBAAiB,EAAE,KAAK;QACxB,UAAU,EAAE,cAAc;QAC1B,SAAS,EAAE,CAAC,OAAO,CAAC;QACpB,UAAU,EAAE;YACV,WAAW,EAAE,SAAS,EAAE,OAAO,EAAE,eAAe;YAChD,OAAO,EAAE,OAAO,EAAE,oBAAoB,EAAE,wBAAwB;SACjE;QACD,UAAU,EAAE,cAAc;QAC1B,WAAW,EAAE;YACX,WAAW,EAAE,MAAM,EAAE,wBAAwB,EAAE,MAAM,EAAE,SAAS,EAAE,MAAM,EAAE,OAAO,EAAE,MAAM;YACzF,eAAe,EAAE,MAAM,EAAE,oBAAoB,EAAE,MAAM;YACrD,wFAAwF;YACxF,8BAA8B;YAC9B,OAAO,EAAE,MAAM,EAAE,OAAO,EAAE,MAAM;SACjC;KACF;IACD,kFAAkF;IAClF,QAAQ,EAAE;QACR,EAAE,EAAE,QAAQ;QACZ,KAAK,EAAE,sDAAsD;QAC7D,KAAK,EAAE,CAAC;QACR,KAAK,EAAE,oBAAoB;QAC3B,QAAQ,EAAE,eAAe;QACzB,YAAY,EAAE,MAAM;QACpB,UAAU,EAAE,QAAQ;QACpB,UAAU,EAAE,WAAW;QACvB,iBAAiB,EAAE,KAAK;QACxB,UAAU,EAAE,cAAc;QAC1B,SAAS,EAAE,CAAC,OAAO,CAAC;QACpB,UAAU,EAAE;YACV,4BAA4B,EAAE,2BAA2B;YACzD,eAAe,EAAE,SAAS,EAAE,iBAAiB;SAC9C;QACD,UAAU,EAAE,cAAc;QAC1B,WAAW,EAAE;YACX,4BAA4B,EAAE,MAAM;YACpC,2BAA2B,EAAE,MAAM;YACnC,eAAe,EAAE,MAAM;YACvB,SAAS,EAAE,MAAM;YACjB,iBAAiB,EAAE,MAAM;SAC1B;KACF;IACD,iGAAiG;IACjG,4FAA4F;IAC5F,6FAA6F;IAC7F,kGAAkG;IAClG,kGAAkG;IAClG,8FAA8F;IAC9F,6CAA6C;IAC7C,iGAAiG;IACjG,iGAAiG;IACjG,oEAAoE;IACpE,kGAAkG;IAClG,gFAAgF;IAChF,OAAO,EAAE;QACP,EAAE,EAAE,OAAO;QACX,KAAK,EAAE,8CAA8C;QACrD,KAAK,EAAE,CAAC;QACR,KAAK,EAAE,eAAe;QACtB,QAAQ,EAAE,UAAU;QACpB,YAAY,EAAE,MAAM;QACpB,UAAU,EAAE,QAAQ,EAAE,6DAA6D;QACnF,UAAU,EAAE,WAAW;QACvB,iBAAiB,EAAE,IAAI,EAAE,+DAA+D;QACxF,UAAU,EAAE,wBAAwB;QACpC,SAAS,EAAE,CAAC,OAAO,CAAC;QACpB,UAAU,EAAE,CAAC,OAAO,EAAE,cAAc,EAAE,YAAY,EAAE,SAAS,CAAC;QAC9D,UAAU,EAAE,wBAAwB;QACpC,WAAW,EAAE;YACX,OAAO,EAAE,MAAM;YACf,cAAc,EAAE,MAAM;YACtB,YAAY,EAAE,MAAM;YACpB,SAAS,EAAE,MAAM;SAClB;KACF;IACD,gGAAgG;IAChG,mGAAmG;IACnG,iGAAiG;IACjG,gGAAgG;IAChG,4EAA4E;IAC5E,OAAO,EAAE;QACP,EAAE,EAAE,OAAO;QACX,KAAK,EAAE,8CAA8C;QACrD,KAAK,EAAE,CAAC;QACR,KAAK,EAAE,eAAe;QACtB,QAAQ,EAAE,eAAe;QACzB,YAAY,EAAE,MAAM;QACpB,UAAU,EAAE,QAAQ;QACpB,UAAU,EAAE,WAAW;QACvB,iBAAiB,EAAE,IAAI,EAAE,mEAAmE;QAC5F,UAAU,EAAE,wBAAwB;QACpC,SAAS,EAAE,CAAC,OAAO,CAAC;QACpB,UAAU,EAAE,CAAC,cAAc,EAAE,YAAY,EAAE,SAAS,CAAC;QACrD,UAAU,EAAE,wBAAwB;QACpC,WAAW,EAAE;YACX,cAAc,EAAE,MAAM;YACtB,YAAY,EAAE,MAAM;YACpB,SAAS,EAAE,MAAM;SAClB;KACF;IACD,+FAA+F;IAC/F,4FAA4F;IAC5F,8FAA8F;IAC9F,iGAAiG;IACjG,sFAAsF;IACtF,+FAA+F;IAC/F,iDAAiD;IACjD,OAAO,EAAE;QACP,EAAE,EAAE,OAAO;QACX,KAAK,EAAE,4CAA4C;QACnD,KAAK,EAAE,CAAC;QACR,KAAK,EAAE,aAAa;QACpB,QAAQ,EAAE,UAAU;QACpB,YAAY,EAAE,MAAM;QACpB,UAAU,EAAE,QAAQ;QACpB,UAAU,EAAE,UAAU;QACtB,iBAAiB,EAAE,IAAI;QACvB,UAAU,EAAE,kBAAkB;QAC9B,SAAS,EAAE,CAAC,OAAO,CAAC;QACpB,UAAU,EAAE,CAAC,qBAAqB,CAAC;QACnC,UAAU,EAAE,EAAE,EAAE,2EAA2E;QAC3F,WAAW,EAAE;YACX,qBAAqB,EAAE,MAAM,EAAE,oDAAoD;SACpF;KACF;IACD,8FAA8F;IAC9F,4FAA4F;IAC5F,4FAA4F;IAC5F,wCAAwC;IACxC,OAAO,EAAE;QACP,EAAE,EAAE,OAAO;QACX,KAAK,EAAE,4CAA4C;QACnD,KAAK,EAAE,CAAC;QACR,KAAK,EAAE,aAAa;QACpB,QAAQ,EAAE,eAAe;QACzB,YAAY,EAAE,MAAM;QACpB,UAAU,EAAE,QAAQ;QACpB,UAAU,EAAE,UAAU;QACtB,iBAAiB,EAAE,IAAI;QACvB,UAAU,EAAE,kBAAkB;QAC9B,SAAS,EAAE,CAAC,OAAO,CAAC;QACpB,UAAU,EAAE,CAAC,qBAAqB,CAAC;QACnC,UAAU,EAAE,EAAE,EAAE,qEAAqE;QACrF,WAAW,EAAE;YACX,qBAAqB,EAAE,MAAM;SAC9B;KACF;IACD,6FAA6F;IAC7F,kGAAkG;IAClG,mGAAmG;IACnG,kGAAkG;IAClG,yDAAyD;IACzD,QAAQ,EAAE;QACR,EAAE,EAAE,QAAQ;QACZ,KAAK,EAAE,+CAA+C;QACtD,KAAK,EAAE,CAAC;QACR,KAAK,EAAE,eAAe;QACtB,QAAQ,EAAE,eAAe;QACzB,YAAY,EAAE,MAAM;QACpB,UAAU,EAAE,QAAQ;QACpB,UAAU,EAAE,SAAS;QACrB,iBAAiB,EAAE,KAAK,EAAE,sEAAsE;QAChG,UAAU,EAAE,kBAAkB;QAC9B,SAAS,EAAE,CAAC,OAAO,EAAE,MAAM,CAAC;QAC5B,UAAU,EAAE,CAAC,cAAc,CAAC;QAC5B,UAAU,EAAE,CAAC,eAAe,EAAE,eAAe,EAAE,eAAe,CAAC;QAC/D,WAAW,EAAE;YACX,cAAc,EAAE,KAAK;SACtB;KACF;CACF,CAAC"}

package/dist/verify-example.d.ts

@@ -0,0 +1,27 @@
+import { type DesignFacets } from "./embedding.js";
+export interface ExampleExpectation {
+    target: string;
+    taskModel?: string;
+    itemType?: string;
+    dimension?: string;
+    standard?: string;
+}
+export interface VerifyResult {
+    ok: boolean;
+    blocking: string[];
+    advisory: string[];
+    facets: DesignFacets;
+}
+/**
+ * Verify one example against its declared expectation. Pure: the caller supplies the compile result
+ * (`errors`, `data`) — parsing/compiling lives at the call site (the parser is not a core dep). Pass
+ * `data` (a composed program from `compiler.compile`) for the authoritative signature; `code` alone
+ * falls back to the source-derived signature.
+ */
+export declare function verifyExample(input: {
+    errors: any[];
+    data?: any;
+    code?: string;
+    expect: ExampleExpectation;
+}): VerifyResult;
+//# sourceMappingURL=verify-example.d.ts.map

package/dist/verify-example.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"verify-example.d.ts","sourceRoot":"","sources":["../src/verify-example.ts"],"names":[],"mappings":"AAeA,OAAO,EAAgD,KAAK,YAAY,EAAE,MAAM,gBAAgB,CAAC;AAGjG,MAAM,WAAW,kBAAkB;IACjC,MAAM,EAAE,MAAM,CAAC;IACf,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,QAAQ,CAAC,EAAE,MAAM,CAAC;CACnB;AAED,MAAM,WAAW,YAAY;IAC3B,EAAE,EAAE,OAAO,CAAC;IACZ,QAAQ,EAAE,MAAM,EAAE,CAAC;IACnB,QAAQ,EAAE,MAAM,EAAE,CAAC;IACnB,MAAM,EAAE,YAAY,CAAC;CACtB;AAQD;;;;;GAKG;AACH,wBAAgB,aAAa,CAAC,KAAK,EAAE;IACnC,MAAM,EAAE,GAAG,EAAE,CAAC;IACd,IAAI,CAAC,EAAE,GAAG,CAAC;IACX,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,MAAM,EAAE,kBAAkB,CAAC;CAC5B,GAAG,YAAY,CAyDf"}

package/dist/verify-example.js

@@ -0,0 +1,82 @@
+// SPDX-License-Identifier: MIT
+// A verification GATE for RAG example data. It does NOT map prompt→code (that stays with the
+// generator + human refinement); it inspects whatever code an example already carries and checks
+// only the machine-decidable properties, so it composes with a human-in-the-loop workflow.
+//
+// Two tiers:
+//   - BLOCKING  — objective failures that must keep an example out of the corpus: it must compile
+//                 clean, and its composed design signature (target / item type / task model /
+//                 dimension / standard) must match the example's declared intent. Task-model checks
+//                 read ./targets.ts (the single source of truth), NOT instructions.md — so the gate
+//                 is independent of the docs a bootstrap run was seeded from.
+//   - ADVISORY  — compiler warnings (thin distractor pool, A↔B giveaway, above-grade reading) that a
+//                 human should judge, surfaced but not fatal.
+//
+// The gate guards BOTH sides of the human: run it after bootstrap AND after hand-refinement.
+import { buildSignatureTags, buildSignatureFromSource } from "./embedding.js";
+import { TARGETS_DATA, taskModelNumber } from "./targets.js";
+// Accept "tm3" or "3"; return both the table key and the bare number.
+function normalizeTaskModel(s) {
+    const num = String(s).toLowerCase().replace(/^tm/, "");
+    return { key: `tm${num}`, num };
+}
+/**
+ * Verify one example against its declared expectation. Pure: the caller supplies the compile result
+ * (`errors`, `data`) — parsing/compiling lives at the call site (the parser is not a core dep). Pass
+ * `data` (a composed program from `compiler.compile`) for the authoritative signature; `code` alone
+ * falls back to the source-derived signature.
+ */
+export function verifyExample(input) {
+    const { errors, data, code, expect } = input;
+    const blocking = [];
+    const advisory = [];
+    // 1. Must compile clean — any hard error is blocking.
+    for (const e of errors || [])
+        blocking.push(`compile error: ${e?.message ?? String(e)}`);
+    // 2. Signature — prefer the composed data; fall back to source.
+    const sig = data != null ? buildSignatureTags(data) : buildSignatureFromSource(code || "");
+    const facets = sig.facets;
+    const target = expect.target;
+    const profile = TARGETS_DATA[target];
+    if (!profile) {
+        blocking.push(`unknown expected target '${target}'`);
+        return { ok: blocking.length === 0, blocking, advisory, facets };
+    }
+    // Resolve the expected item type / task-model number from whatever the expectation provided, and
+    // cross-check them when both are given.
+    const expTm = expect.taskModel ? normalizeTaskModel(expect.taskModel) : undefined;
+    let expItemType = expect.itemType;
+    if (!expItemType && expTm)
+        expItemType = profile.taskModels[expTm.key];
+    const expTmNum = expTm?.num ?? (expItemType ? taskModelNumber(target, expItemType) : undefined);
+    if (expect.itemType && expTm && profile.taskModels[expTm.key] !== expect.itemType) {
+        blocking.push(`expectation is inconsistent: ${expTm.key} is '${profile.taskModels[expTm.key] ?? "n/a"}' for ${target}, not '${expect.itemType}'`);
+    }
+    if (expTm && !profile.taskModels[expTm.key]) {
+        blocking.push(`task model '${expTm.key}' is not available for target ${target}`);
+    }
+    // 3. Composed signature must match the declared intent.
+    if (facets.target !== target) {
+        blocking.push(`target: expected ${target}, composed ${facets.target ?? "none"}`);
+    }
+    const gotTypes = facets.itemTypes || [];
+    if (expItemType && !(gotTypes.length === 1 && gotTypes[0] === expItemType)) {
+        blocking.push(`item type: expected exactly [${expItemType}], composed [${gotTypes.join(", ") || "none"}]`);
+    }
+    if (expTmNum && !(facets.taskModels || []).includes(expTmNum)) {
+        blocking.push(`task model: expected tm${expTmNum}, composed task-model(s) [${(facets.taskModels || []).join(", ") || "none"}]`);
+    }
+    if (expect.dimension && !(facets.dimensions || []).includes(expect.dimension)) {
+        blocking.push(`dimension: expected ${expect.dimension}, composed [${(facets.dimensions || []).join(", ") || "none"}]`);
+    }
+    if (expect.standard && !(facets.standards || []).includes(expect.standard)) {
+        blocking.push(`standard: expected ${expect.standard}, composed [${(facets.standards || []).join(", ") || "none"}]`);
+    }
+    // 4. Compiler warnings → advisory (a human decides). Warnings ride on each composed item.
+    const items = data == null ? [] : Array.isArray(data) ? data : data.kind === "items" ? data.items : [data];
+    for (const it of items)
+        for (const w of (it && it.warnings) || [])
+            advisory.push(w);
+    return { ok: blocking.length === 0, blocking, advisory, facets };
+}
+//# sourceMappingURL=verify-example.js.map

package/dist/verify-example.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"verify-example.js","sourceRoot":"","sources":["../src/verify-example.ts"],"names":[],"mappings":"AAAA,+BAA+B;AAC/B,6FAA6F;AAC7F,iGAAiG;AACjG,2FAA2F;AAC3F,EAAE;AACF,aAAa;AACb,kGAAkG;AAClG,8FAA8F;AAC9F,oGAAoG;AACpG,oGAAoG;AACpG,8EAA8E;AAC9E,qGAAqG;AACrG,8DAA8D;AAC9D,EAAE;AACF,6FAA6F;AAC7F,OAAO,EAAE,kBAAkB,EAAE,wBAAwB,EAAqB,MAAM,gBAAgB,CAAC;AACjG,OAAO,EAAE,YAAY,EAAE,eAAe,EAAE,MAAM,cAAc,CAAC;AAiB7D,sEAAsE;AACtE,SAAS,kBAAkB,CAAC,CAAS;IACnC,MAAM,GAAG,GAAG,MAAM,CAAC,CAAC,CAAC,CAAC,WAAW,EAAE,CAAC,OAAO,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC;IACvD,OAAO,EAAE,GAAG,EAAE,KAAK,GAAG,EAAE,EAAE,GAAG,EAAE,CAAC;AAClC,CAAC;AAED;;;;;GAKG;AACH,MAAM,UAAU,aAAa,CAAC,KAK7B;IACC,MAAM,EAAE,MAAM,EAAE,IAAI,EAAE,IAAI,EAAE,MAAM,EAAE,GAAG,KAAK,CAAC;IAC7C,MAAM,QAAQ,GAAa,EAAE,CAAC;IAC9B,MAAM,QAAQ,GAAa,EAAE,CAAC;IAE9B,sDAAsD;IACtD,KAAK,MAAM,CAAC,IAAI,MAAM,IAAI,EAAE;QAAE,QAAQ,CAAC,IAAI,CAAC,kBAAkB,CAAC,EAAE,OAAO,IAAI,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;IAEzF,gEAAgE;IAChE,MAAM,GAAG,GAAG,IAAI,IAAI,IAAI,CAAC,CAAC,CAAC,kBAAkB,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,wBAAwB,CAAC,IAAI,IAAI,EAAE,CAAC,CAAC;IAC3F,MAAM,MAAM,GAAG,GAAG,CAAC,MAAM,CAAC;IAE1B,MAAM,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;IAC7B,MAAM,OAAO,GAAG,YAAY,CAAC,MAAM,CAAC,CAAC;IACrC,IAAI,CAAC,OAAO,EAAE,CAAC;QACb,QAAQ,CAAC,IAAI,CAAC,4BAA4B,MAAM,GAAG,CAAC,CAAC;QACrD,OAAO,EAAE,EAAE,EAAE,QAAQ,CAAC,MAAM,KAAK,CAAC,EAAE,QAAQ,EAAE,QAAQ,EAAE,MAAM,EAAE,CAAC;IACnE,CAAC;IAED,iGAAiG;IACjG,wCAAwC;IACxC,MAAM,KAAK,GAAG,MAAM,CAAC,SAAS,CAAC,CAAC,CAAC,kBAAkB,CAAC,MAAM,CAAC,SAAS,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IAClF,IAAI,WAAW,GAAG,MAAM,CAAC,QAAQ,CAAC;IAClC,IAAI,CAAC,WAAW,IAAI,KAAK;QAAE,WAAW,GAAG,OAAO,CAAC,UAAU,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC;IACvE,MAAM,QAAQ,GAAG,KAAK,EAAE,GAAG,IAAI,CAAC,WAAW,CAAC,CAAC,CAAC,eAAe,CAAC,MAAM,EAAE,WAAW,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC;IAChG,IAAI,MAAM,CAAC,QAAQ,IAAI,KAAK,IAAI,OAAO,CAAC,UAAU,CAAC,KAAK,CAAC,GAAG,CAAC,KAAK,MAAM,CAAC,QAAQ,EAAE,CAAC;QAClF,QAAQ,CAAC,IAAI,CACX,gCAAgC,KAAK,CAAC,GAAG,QAAQ,OAAO,CAAC,UAAU,CAAC,KAAK,CAAC,GAAG,CAAC,IAAI,KAAK,SAAS,MAAM,UAAU,MAAM,CAAC,QAAQ,GAAG,CACnI,CAAC;IACJ,CAAC;IACD,IAAI,KAAK,IAAI,CAAC,OAAO,CAAC,UAAU,CAAC,KAAK,CAAC,GAAG,CAAC,EAAE,CAAC;QAC5C,QAAQ,CAAC,IAAI,CAAC,eAAe,KAAK,CAAC,GAAG,iCAAiC,MAAM,EAAE,CAAC,CAAC;IACnF,CAAC;IAED,wDAAwD;IACxD,IAAI,MAAM,CAAC,MAAM,KAAK,MAAM,EAAE,CAAC;QAC7B,QAAQ,CAAC,IAAI,CAAC,oBAAoB,MAAM,cAAc,MAAM,CAAC,MAAM,IAAI,MAAM,EAAE,CAAC,CAAC;IACnF,CAAC;IACD,MAAM,QAAQ,GAAG,MAAM,CAAC,SAAS,IAAI,EAAE,CAAC;IACxC,IAAI,WAAW,IAAI,CAAC,CAAC,QAAQ,CAAC,MAAM,KAAK,CAAC,IAAI,QAAQ,CAAC,CAAC,CAAC,KAAK,WAAW,CAAC,EAAE,CAAC;QAC3E,QAAQ,CAAC,IAAI,CAAC,gCAAgC,WAAW,gBAAgB,QAAQ,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,MAAM,GAAG,CAAC,CAAC;IAC7G,CAAC;IACD,IAAI,QAAQ,IAAI,CAAC,CAAC,MAAM,CAAC,UAAU,IAAI,EAAE,CAAC,CAAC,QAAQ,CAAC,QAAQ,CAAC,EAAE,CAAC;QAC9D,QAAQ,CAAC,IAAI,CAAC,0BAA0B,QAAQ,6BAA6B,CAAC,MAAM,CAAC,UAAU,IAAI,EAAE,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,MAAM,GAAG,CAAC,CAAC;IAClI,CAAC;IACD,IAAI,MAAM,CAAC,SAAS,IAAI,CAAC,CAAC,MAAM,CAAC,UAAU,IAAI,EAAE,CAAC,CAAC,QAAQ,CAAC,MAAM,CAAC,SAAS,CAAC,EAAE,CAAC;QAC9E,QAAQ,CAAC,IAAI,CAAC,uBAAuB,MAAM,CAAC,SAAS,eAAe,CAAC,MAAM,CAAC,UAAU,IAAI,EAAE,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,MAAM,GAAG,CAAC,CAAC;IACzH,CAAC;IACD,IAAI,MAAM,CAAC,QAAQ,IAAI,CAAC,CAAC,MAAM,CAAC,SAAS,IAAI,EAAE,CAAC,CAAC,QAAQ,CAAC,MAAM,CAAC,QAAQ,CAAC,EAAE,CAAC;QAC3E,QAAQ,CAAC,IAAI,CAAC,sBAAsB,MAAM,CAAC,QAAQ,eAAe,CAAC,MAAM,CAAC,SAAS,IAAI,EAAE,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,MAAM,GAAG,CAAC,CAAC;IACtH,CAAC;IAED,0FAA0F;IAC1F,MAAM,KAAK,GAAG,IAAI,IAAI,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,KAAK,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;IAC3G,KAAK,MAAM,EAAE,IAAI,KAAK;QAAE,KAAK,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,CAAC,QAAQ,CAAC,IAAI,EAAE;YAAE,QAAQ,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAEpF,OAAO,EAAE,EAAE,EAAE,QAAQ,CAAC,MAAM,KAAK,CAAC,EAAE,QAAQ,EAAE,QAAQ,EAAE,MAAM,EAAE,CAAC;AACnE,CAAC"}

package/package.json

@@ -3,7 +3,7 @@
     "access": "public"
   },
   "name": "@graffiticode/l0175",
-  "version": "0.2.0",
+  "version": "0.3.1",
   "description": "L0175 — a Graffiticode dialect (hello/image/theme/print) inheriting @graffiticode/l0000",
   "type": "module",
   "main": "./dist/index.js",
@@ -25,7 +25,7 @@
     "lint": "eslint src tools"
   },
   "dependencies": {
-    "@graffiticode/l0000": "^0.1.2"
+    "@graffiticode/l0000": "^0.1.3"
   },
   "devDependencies": {
     "spec-md": "^3.1.0"

package/spec/#template.gc#

@@ -0,0 +1,20 @@
+passage "Title"
+  type literary
+  lines [
+    "Line one."
+  ] claims []
+  evidence []
+  outcomes [] {}..
+Read the sentence: "The aqueduct carried water across long distances using gentle slopes." What does the word aqueduct most likely mean?
+
+A. a channel built to carry water ✓
+
+B. a kind of paved road
+
+C. a boat that carries cargo
+
+D. a tall stone tower
+
+Answer key: Answer — A
+
+Correct inference: a channel built to carry water