eklavya 1.30.3 → 1.30.4

package/dist/assets/tutor/references/grading.md

@@ -41,6 +41,12 @@ Within the cap, still grade honestly:
 | 1 | picked a distractor built on a misconception |
 | 0 | "Other" with *I don't know* (`outcome: dont_know` — **teach it**), or a decline (`outcome: declined`) |
 
+**A defensible pick is a correct answer.** If the option they chose also
+answers the stem — or they argue convincingly that it does — the question was
+flawed, not the learner. Grade it 4 as if it were the key, say plainly that the
+question had two right answers, and never grade them down for your own
+ambiguity.
+
 **If they want to explain, let them, and say so.** Someone who picks "Other" and
 types a real answer has just given you better evidence than the multiple choice
 could. Grade that as the free answer it is — **omit `format`**, and the cap does

package/dist/assets/tutor/references/writing-mcq.md

@@ -43,6 +43,14 @@ learner can dismiss without thinking is a free point, and it teaches nothing.
 When three that good will not come, the stem is too vague to have a near-miss:
 rewrite the stem and the distractors follow.
 
+**Exactly one option answers the stem.** Plausible is not the same as
+defensible: a distractor must be believable *and wrong as an answer to this
+question*. Row two is where this breaks — a true statement answers most "why"
+stems well enough, and a learner who picks it is marked wrong for being right.
+Before asking, read each distractor as if it were the key: if a senior reviewer
+could argue for it, narrow the stem ("why *here*", "what does *this line*
+prevent") until it cannot, or replace it with a different row.
+
 **4. One clause of `description` per option.** This is where a near-miss earns
 its place — the sentence that makes the wrong answer tempting.
 
@@ -90,6 +98,7 @@ telling you which part to rebuild.
   grammar. A visibly longer or more careful option reads as the correct one, and
   gets picked without engaging — the same leak as always answering first.
 - Each of the three wrong options came from a different row of the table above.
+- Only the correct option answers the stem; no expert could argue for another.
 - The correct option sits at `answer_position`.
 - The tool renders the labels, so the stem does not number them.
 - The stem is the question and nothing else — the dials are in the developer's

package/dist/hooks/checkpoint-quiz.js

@@ -149,7 +149,7 @@ Concept: ${row.concept}
 
 Do exactly this, then get straight back to the task:
   1. get_session_quiz_plan with max: 1 and ignore_cooldown: true (the pacing is already decided -- this hook is the cooldown).
-  2. Ask that ONE question with AskUserQuestion: four options, one correct, three plausible, and put the correct one in the slot answer_position names.
+  2. Ask that ONE question with AskUserQuestion: four options, exactly one correct, three plausible but wrong for this stem, and put the correct one in the slot answer_position names.
      ${attributionRule()}
   3. Grade it with record_attempt: format "mcq", the labels in "options", the stem alone in "question".
   4. Tell them the verdict: right, or wrong and what the right answer is, with one line of why. Never skip this -- an answer with no verdict teaches nothing.

package/dist/plugin/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "eklavya",
   "displayName": "Eklavya",
-  "version": "1.30.3",
+  "version": "1.30.4",
   "description": "Learn while your agent works. Turns coding-agent generation time into adaptive, Socratic learning grounded in the code being written.",
   "author": {
     "name": "Ajay Kumar"

package/dist/plugin/agents/tutor.md

@@ -56,7 +56,8 @@ replying "A" without reading.
 a blank, not a decline: grade 0 with `outcome: "dont_know"`, then teach.
 
 Same rules as the tutor skill's `skills/tutor/references/writing-mcq.md`: four
-options, three plausible distractors, the correct one at `answer_position`, the
+options, exactly one that answers the stem, three plausible distractors that do
+not, the correct one at `answer_position`, the
 stem alone in `record_attempt`'s `question`, the labels in `options`,
 `format: "mcq"`, and the grade capped at 4. The only thing that changes is who
 draws the box.

package/dist/plugin/skills/tutor/references/grading.md

@@ -41,6 +41,12 @@ Within the cap, still grade honestly:
 | 1 | picked a distractor built on a misconception |
 | 0 | "Other" with *I don't know* (`outcome: dont_know` — **teach it**), or a decline (`outcome: declined`) |
 
+**A defensible pick is a correct answer.** If the option they chose also
+answers the stem — or they argue convincingly that it does — the question was
+flawed, not the learner. Grade it 4 as if it were the key, say plainly that the
+question had two right answers, and never grade them down for your own
+ambiguity.
+
 **If they want to explain, let them, and say so.** Someone who picks "Other" and
 types a real answer has just given you better evidence than the multiple choice
 could. Grade that as the free answer it is — **omit `format`**, and the cap does

package/dist/plugin/skills/tutor/references/writing-mcq.md

@@ -43,6 +43,14 @@ learner can dismiss without thinking is a free point, and it teaches nothing.
 When three that good will not come, the stem is too vague to have a near-miss:
 rewrite the stem and the distractors follow.
 
+**Exactly one option answers the stem.** Plausible is not the same as
+defensible: a distractor must be believable *and wrong as an answer to this
+question*. Row two is where this breaks — a true statement answers most "why"
+stems well enough, and a learner who picks it is marked wrong for being right.
+Before asking, read each distractor as if it were the key: if a senior reviewer
+could argue for it, narrow the stem ("why *here*", "what does *this line*
+prevent") until it cannot, or replace it with a different row.
+
 **4. One clause of `description` per option.** This is where a near-miss earns
 its place — the sentence that makes the wrong answer tempting.
 
@@ -90,6 +98,7 @@ telling you which part to rebuild.
   grammar. A visibly longer or more careful option reads as the correct one, and
   gets picked without engaging — the same leak as always answering first.
 - Each of the three wrong options came from a different row of the table above.
+- Only the correct option answers the stem; no expert could argue for another.
 - The correct option sits at `answer_position`.
 - The tool renders the labels, so the stem does not number them.
 - The stem is the question and nothing else — the dials are in the developer's

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "eklavya",
-  "version": "1.30.3",
+  "version": "1.30.4",
   "description": "Learn while your agent works — local knowledge graph, spaced repetition, and quiz gating for agent-assisted development.",
   "license": "MIT",
   "repository": {