tutor-resume

name

tutor-resume

description

Pick up a paused tutoring session. Finds the user's active in-progress topic, briefly reorients them (where they are, what's acquired, what's coming), and hands the dialogue back to the active persona. If multiple topics are in progress, asks which one. Invoke when you're returning after a break. Triggers: '/tutor-resume', 'continue', 'pick up where I left off', 'back to the tutor', 'resume tutoring'.

Tutor Resume

The "pick up where you left off" skill. After a break — minutes, hours, or days — the user invokes this to be re-oriented into their active topic without having to remember the slug or re-invoke /tutor-start.

This skill is read-mostly: it queries state to find the active topic, prints an orientation message, and hands off to the persona. It does NOT charge cycles (the user already paid when they originally invoked /tutor-start on this topic; resume is free).

---

Step 0 — Defensive setup

STATE="$CLAUDE_PLUGIN_ROOT/scripts/state.sh"
bash "$STATE" init >/dev/null
bash "$STATE" refill-cycles >/dev/null 2>&1 || true

---

Step 1 — Find active topics

ACTIVE_SLUGS=$(bash "$STATE" get '[.topics | to_entries[] | select(.value.status == "in_progress") | .key]')
ACTIVE_COUNT=$(echo "$ACTIVE_SLUGS" | jq 'length')

Branch: no active topics

If ACTIVE_COUNT is 0:

"No topics in progress. Run /tutor-start <subject> to begin one, or /tutor-status to see your completed topics."

Exit cleanly.

Branch: one active topic

If ACTIVE_COUNT is 1, use that slug. Skip to Step 2.

SLUG=$(echo "$ACTIVE_SLUGS" | jq -r '.[0]')

Branch: multiple active topics

If ACTIVE_COUNT > 1, list them with brief progress info and ask the user to pick.

echo "You have $ACTIVE_COUNT active topics:"
echo
echo "$ACTIVE_SLUGS" | jq -r '.[]' | nl -w1 -s'. ' | while read -r line; do
  # line is "1. slug-name"
  num=$(echo "$line" | cut -d. -f1)
  slug=$(echo "$line" | cut -d'.' -f2- | sed 's/^ //')
  # Display the human name (display_name set at /tutor-start time) when
  # present; fall back to the slug-key for topics created before v0.5.0.
  display=$(bash "$STATE" get ".topics[\"$slug\"].display_name // \"$slug\"")
  acquired=$(bash "$STATE" get ".topics[\"$slug\"].concepts | map(select(.status == \"acquired\")) | length")
  total=$(bash "$STATE" get ".topics[\"$slug\"].concepts | length")
  current_name=$(bash "$STATE" get ".topics[\"$slug\"].concepts[\"$(bash "$STATE" get ".topics[\"$slug\"].current_concept_index")\"].name // \"?\"")
  printf "  %s. %-30s (concept %s/%s: %s)\n" "$num" "$display" "$((acquired + 1))" "$total" "$current_name"
done
echo
echo "Which one are you resuming? (Type the name or the number.)"

Wait for the user's answer. Accept either:

• A 1-based number (e.g., "2") — most common
• A substring of any display_name or slug — match case-insensitively
• The exact slug-key

Validate the input. If invalid: "I don't see that as one of the active topics. Try again." Re-prompt.

---

Step 2 — Read the topic's state for orientation

CONCEPTS=$(bash "$STATE" get ".topics[\"$SLUG\"].concepts")
CURRENT_IDX=$(bash "$STATE" get ".topics[\"$SLUG\"].current_concept_index")
CALIBRATION=$(bash "$STATE" get ".topics[\"$SLUG\"].calibration")
# display_name with slug-key fallback for topics created before v0.5.0.
DISPLAY=$(bash "$STATE" get ".topics[\"$SLUG\"].display_name // \"$SLUG\"")
TOTAL=$(echo "$CONCEPTS" | jq 'length')
ACQUIRED_COUNT=$(echo "$CONCEPTS" | jq '[.[] | select(.status == "acquired")] | length')

# Current concept name (1-based for display)
CURRENT_NAME=$(echo "$CONCEPTS" | jq -r ".[$CURRENT_IDX].name")
CURRENT_POS=$((CURRENT_IDX + 1))

# Comma-separated names of acquired concepts (in order)
ACQUIRED_NAMES=$(echo "$CONCEPTS" | jq -r '[.[] | select(.status == "acquired") | .name] | join(", ")')

# Comma-separated names of upcoming pending concepts (in order, AFTER current)
UPCOMING_NAMES=$(echo "$CONCEPTS" | jq -r --argjson i $CURRENT_IDX \
  '[.[($i + 1):] | .[] | select(.status == "pending") | .name] | join(" → ")')

---

Step 3 — Print the orientation

The orientation is brief — 3-4 lines max. The user is RESUMING, not starting; they don't need a tutorial on what the path looks like (that's /tutor-path).

echo "Resuming $DISPLAY."
echo
echo "You're at concept $CURRENT_POS/$TOTAL: $CURRENT_NAME"
if [ -n "$ACQUIRED_NAMES" ]; then
  echo "Already acquired: $ACQUIRED_NAMES"
fi
if [ -n "$UPCOMING_NAMES" ] && [ "$UPCOMING_NAMES" != "" ]; then
  echo "Up next: $UPCOMING_NAMES"
fi
echo

Example output:

Resuming Python decorators.

You're at concept 3/6: wrapper functions
Already acquired: closures, first-class functions
Up next: @ syntax → decorators with arguments → real-world patterns

---

Step 4 — Hand off to the persona

This is the hand-off cue. The active persona (Echo / Cipher / Vex via output style) picks up the dialogue from here.

The hand-off message depends on the calibration branch the user was on when the topic was started:

For intermediate calibration (productive-failure path)

"Picking up at $CURRENT_NAME. Match what we were doing before — pose a fresh problem on this concept (productive-failure style) and let the user attempt. If you can sense they need scaffolding from how long the pause has been, drop down a hint rung; otherwise keep the difficulty where it was."

For novice calibration ("I do / We do / You do" path)

"Picking up at $CURRENT_NAME. Match what we were doing before — worked example or guided practice as appropriate. If you can sense they need a refresher, recap in one sentence before the next question; otherwise keep moving."

The skill ends here. The persona takes over.

---

Hard rules

1. No cycles charge. Resume is free; the user already paid when they first invoked /tutor-start on this topic. The cycles_paid: true flag in state is the receipt.
2. Don't re-do calibration. The user calibrated when they started the topic. Resume picks up where they left off; it doesn't re-test.
3. No path printing. The orientation message doesn't render the full learning path. The user can invoke /tutor-path if they want that.
4. Brief orientation. 3-4 lines, then hand off. Long orientations break the resume feel — the user wants to continue working, not read a status report.
5. Don't auto-pick on ambiguity. When multiple topics are in progress, ask. Auto-picking the most-recent is a guess; the user might want the older one.

---

Edge cases

• current_concept_index is past the array length (could happen if /tutor-done was almost called but failed) → Treat as "at the last concept" and continue. Don't crash.
• Active topic with no concepts (a failed /tutor-start) → Print: "$SLUG has no concepts — looks like the path generation failed. Try /tutor-start $SLUG to regenerate." Exit.
• User picks a slug not in the active list when prompted → Re-prompt with the same list. Don't fall through to /tutor-start semantics.
• All concepts are already acquired but status is still in_progress → Tell user: "All concepts on this path are marked acquired. You can run /tutor-done $SLUG to close it out." Exit.