refactor: remove all mode references from hooks, skills, and specs

- Remove mode reading from inject-context.sh and subagent-context.sh
- Remove "Modes of Operation" section from ai-dlc-fundamentals
- Delete ai-dlc-mode-selection skill entirely
- Skip integrator for single-unit intents (reviewer already validated)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: jwaldrip

plugin/hooks/inject-context.sh

@@ -318,19 +318,9 @@ if [ "$STATUS" = "complete" ]; then
   exit 0
 fi
 
-# Read mode from intent.md frontmatter (mode lives on the intent, not hats)
-INTENT_SLUG_EARLY="${KEEP_VALUES[intent-slug]:-}"
-MODE=""
-if [ -n "$INTENT_SLUG_EARLY" ] && [ -f ".ai-dlc/${INTENT_SLUG_EARLY}/intent.md" ]; then
-  MODE=$(han parse yaml mode -r --default OHOTL < ".ai-dlc/${INTENT_SLUG_EARLY}/intent.md" 2>/dev/null || echo "OHOTL")
-else
-  # Try parsing mode from iteration.json if intent.md not available yet
-  MODE=$(echo "$ITERATION_JSON" | han parse json mode -r --default OHOTL 2>/dev/null || echo "OHOTL")
-fi
-
 echo "## AI-DLC Context"
 echo ""
-echo "**Iteration:** $ITERATION | **Hat:** $HAT | **Mode:** $MODE | **Workflow:** $WORKFLOW_NAME ($WORKFLOW_HATS_STR)"
+echo "**Iteration:** $ITERATION | **Hat:** $HAT | **Workflow:** $WORKFLOW_NAME ($WORKFLOW_HATS_STR)"
 echo ""
 
 # Inject provider context

plugin/hooks/subagent-context.sh

@@ -176,12 +176,11 @@ if [ -z "${CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS:-}" ]; then
   if [ -n "$HAT_FILE" ] && [ -f "$HAT_FILE" ]; then
     # Parse frontmatter
     NAME=$(han parse yaml name -r --default "" < "$HAT_FILE" 2>/dev/null || echo "")
-    MODE=$(han parse yaml mode -r --default "" < "$HAT_FILE" 2>/dev/null || echo "")
 
     # Get content after frontmatter
     INSTRUCTIONS=$(cat "$HAT_FILE" | sed '1,/^---$/d' | sed '1,/^---$/d')
 
-    echo "**${NAME:-$HAT}** (Mode: ${MODE:-HITL})"
+    echo "**${NAME:-$HAT}**"
     echo ""
     if [ -n "$INSTRUCTIONS" ]; then
       echo "$INSTRUCTIONS"
@@ -258,7 +257,7 @@ echo "3. **If on wrong branch** - Switch to correct branch and continue"
 echo "4. **If tests fail** - Fix and retry, don't give up"
 echo "5. **Only declare blocked** after 3+ genuine rescue attempts"
 echo ""
-echo "### Communication (HITL)"
+echo "### Communication"
 echo ""
 echo "**Notify users of important events:**"
 echo ""

plugin/skills/advance/SKILL.md

@@ -111,8 +111,10 @@ READY_COUNT=$(echo "$DAG_SUMMARY" | han parse json readyCount -r)
 
 ```javascript
 if (dagSummary.allComplete) {
-  // ALL UNITS COMPLETE - Check if integrator has run
-  if (!state.integratorComplete) {
+  // ALL UNITS COMPLETE - Check if integrator should run
+  // Skip integrator for single-unit intents (reviewer already validated it)
+  const unitCount = dagSummary.totalCount;
+  if (unitCount > 1 && !state.integratorComplete) {
     // Spawn integrator on the intent branch
     // See Step 2e below
     return spawnIntegrator();

plugin/skills/ai-dlc-fundamentals/SKILL.md

@@ -150,11 +150,11 @@ Elaboration (`/elaborate`) is a separate pre-construction phase. The constructio
 planner → builder → reviewer
 ```
 
-| Hat | Focus | Mode |
-|-----|-------|------|
-| Planner | Plan this iteration | HITL |
-| Builder | Implement to spec | OHOTL |
-| Reviewer | Verify quality | HITL |
+| Hat | Focus |
+|-----|-------|
+| Planner | Plan this iteration |
+| Builder | Implement to spec |
+| Reviewer | Verify quality |
 
 ### Hat Transitions
 
@@ -186,31 +186,6 @@ Investigate the problem space before implementing.
 Gather context, explore options, document findings.
 ```
 
-## Modes of Operation
-
-### HITL - Human In The Loop
-
-Human actively participates in every decision:
-- Elaboration phase (defining what to build)
-- Review phase (approving implementation)
-- Course corrections (when AI goes off track)
-
-### OHOTL - Occasional Human Over The Loop
-
-Human sets direction, AI operates autonomously:
-- Builder phase with clear criteria
-- Human intervenes only when stuck or for review
-- Backpressure enforces quality automatically
-
-### AHOTL - Autonomous Human Over The Loop
-
-AI operates with minimal human involvement:
-- Multiple iterations without human input
-- Human reviews at end or on exception
-- Requires very clear criteria and robust backpressure
-
-*For detailed mode selection guidance, see the `ai-dlc-mode-selection` skill.*
-
 ## State Management
 
 ### Scoped Storage