@longtable/checkpoints 0.1.38 → 0.1.40

package/dist/trigger-classifier.js

@@ -3,8 +3,8 @@ const RULES = [
         family: "meta_decision",
         checkpointKey: "meta_decision",
         stage: "problem_framing",
-        mode: "commit",
-        level: "adaptive_required",
+        mode: "review",
+        level: "recommended",
         cues: [
             /\brename\b/i,
             /\bnaming\b/i,
@@ -15,7 +15,7 @@ const RULES = [
             /\bmake .*authoritative\b/i,
             /이름|명명|리드미|체크포인트 정책|플랫폼 용어|제품 언어/
         ],
-        rationale: "Platform language and checkpoint policy shape future LongTable behavior."
+        rationale: "Platform language and checkpoint policy should be discussed without creating a research-state checkpoint."
     },
     {
         family: "submission",
@@ -156,7 +156,7 @@ const RULES = [
         checkpointKey: "knowledge_gap_probe",
         stage: "problem_framing",
         mode: "explore",
-        level: "adaptive_required",
+        level: "recommended",
         cues: [
             /\bknowledge gap\b/i,
             /\bgap in (my|our|the) knowledge\b/i,
@@ -171,14 +171,14 @@ const RULES = [
             /\bhelp me narrow\b/i,
             /지식의 공백|지식 공백|모르겠|불확실|놓치고 있|빠뜨리고 있/
         ],
-        rationale: "A named knowledge gap should be surfaced before LongTable narrows or closes the problem."
+        rationale: "A named knowledge gap should be surfaced as response-only guidance unless a research commitment is being made."
     },
     {
         family: "review",
         checkpointKey: "tacit_assumption_probe",
         stage: "problem_framing",
         mode: "review",
-        level: "adaptive_required",
+        level: "recommended",
         cues: [
             /\btacit\b/i,
             /\bimplicit\b/i,
@@ -188,7 +188,7 @@ const RULES = [
             /\bhidden constraint\b/i,
             /암묵지|암묵적|전제|가정|사각지대|숨은 제약|암묵/
         ],
-        rationale: "Tacit assumptions can silently shift research responsibility and need an explicit researcher-facing check."
+        rationale: "Tacit assumptions should be made visible without becoming a durable checkpoint unless they block a research commitment."
     },
     {
         family: "exploration",
@@ -275,12 +275,22 @@ function requiresQuestion(level, mode, family) {
         mode === "submit" ||
         family === "meta_decision");
 }
+function looksLikeCommitmentCue(prompt) {
+    return /\b(final|finalize|commit|ship|submit|publish|freeze|settle|decide|lock|record|apply|incorporate)\b/i.test(prompt)
+        || /최종|확정|커밋|제출|투고|고정|결정|기록|반영/.test(prompt);
+}
+function ruleRequiresQuestionForPrompt(rule, prompt) {
+    if (rule.family === "commitment" && !looksLikeCommitmentCue(prompt)) {
+        return false;
+    }
+    return requiresQuestion(rule.level, rule.mode ?? "review", rule.family);
+}
 function pickBestRule(prompt) {
     const candidates = RULES.map((rule) => ({ rule, matches: matchCues(prompt, rule) }))
         .filter((candidate) => candidate.matches.length > 0)
         .sort((a, b) => {
-        const requiredDelta = Number(requiresQuestion(b.rule.level, b.rule.mode ?? "review", b.rule.family)) -
-            Number(requiresQuestion(a.rule.level, a.rule.mode ?? "review", a.rule.family));
+        const requiredDelta = Number(ruleRequiresQuestionForPrompt(b.rule, prompt)) -
+            Number(ruleRequiresQuestionForPrompt(a.rule, prompt));
         if (requiredDelta !== 0)
             return requiredDelta;
         const genericDelta = Number(a.rule.checkpointKey === "research_question_freeze") -
@@ -332,15 +342,40 @@ function fallbackLevel(mode) {
             return "recommended";
     }
 }
+function looksLikeProductOrToolingPrompt(prompt) {
+    return /\b(longlongtable|hook|checkpoint|mcp|agents?|skills?|ux|interface|setup|install|cli|npm|version|global|release|deploy|git|github|readme|docs?|documentation|workflow|package|router|autocomplete|simulation test)\b/i.test(prompt)
+        || /롱테이블|훅|체크포인트|에이전트|스킬|사용성|인터페이스|설치|세팅|글로벌|배포|버전|릴리즈|깃|깃허브|문서화된\s*절차|패키지|라우터|자동완성|시뮬레이션\s*테스트/.test(prompt);
+}
 export function classifyCheckpointTrigger(prompt, options = {}) {
     const normalizedPrompt = prompt.trim();
+    if (looksLikeProductOrToolingPrompt(normalizedPrompt)) {
+        return {
+            signal: {
+                checkpointKey: options.checkpointKey ?? "product_runtime_guidance",
+                baseLevel: "recommended",
+                mode: options.preferredMode ?? options.fallbackMode ?? "review",
+                artifactStakes: options.artifactStakes ?? "private_note",
+                researchStage: options.researchStage ?? "problem_framing",
+                ...(options.unresolvedTensions ? { unresolvedTensions: options.unresolvedTensions } : {}),
+                ...(options.studyContract ? { studyContract: options.studyContract } : {})
+            },
+            family: "advisory",
+            confidence: "high",
+            matchedCues: ["product_tooling"],
+            requiresQuestionBeforeClosure: false,
+            advisoryOnly: true,
+            rationale: ["Product, tooling, and LongTable policy work should use response-only discussion unless the researcher explicitly records a research decision."]
+        };
+    }
     const best = pickBestRule(normalizedPrompt);
     const inferredMode = inferModeFromPlainCue(normalizedPrompt);
     const mode = options.preferredMode ?? best?.rule.mode ?? inferredMode ?? options.fallbackMode ?? "explore";
     const stage = options.researchStage ?? best?.rule.stage ?? fallbackStage(mode);
     const stakes = options.artifactStakes ?? best?.rule.stakes ?? fallbackStakes(mode);
-    const level = best?.rule.level ?? fallbackLevel(mode);
     const family = best?.rule.family ?? (mode === "explore" ? "exploration" : "advisory");
+    const level = family === "commitment" && !looksLikeCommitmentCue(normalizedPrompt)
+        ? "recommended"
+        : best?.rule.level ?? fallbackLevel(mode);
     const matchedCues = best?.matches ?? [];
     const checkpointKey = options.checkpointKey ?? best?.rule.checkpointKey ?? `${mode}_runtime_guidance`;
     const rationale = [

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@longtable/checkpoints",
-  "version": "0.1.38",
+  "version": "0.1.40",
   "private": false,
   "description": "Checkpoint policy engine for LongTable",
   "type": "module",
@@ -21,7 +21,7 @@
     "typecheck": "tsc -p tsconfig.json --noEmit"
   },
   "dependencies": {
-    "@longtable/core": "0.1.38"
+    "@longtable/core": "0.1.40"
   },
   "devDependencies": {
     "typescript": "^5.6.0"