refactor: remove mode selection from elaboration and construction

The HITL/OHOTL/AHOTL mode concept duplicates Claude Code's native
permission system (plan/acceptEdits/bypassPermissions). Users control
oversight by changing Claude's mode, not by baking it into the spec.

Removes Phase 3.5, mode from intent.md frontmatter, mode from
iteration.json, and modeToAgentTeamsMode mapping from Task spawns.

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

Author: jwaldrip

plugin/skills/construct/SKILL.md

@@ -212,9 +212,8 @@ Initialize `iteration.json` from the intent artifacts:
 INTENT_DIR=".ai-dlc/${INTENT_SLUG}"
 INTENT_FILE="$INTENT_DIR/intent.md"
 
-# Read workflow and mode from intent.md frontmatter
+# Read workflow from intent.md frontmatter
 WORKFLOW_NAME=$(han parse yaml workflow -r --default "default" < "$INTENT_FILE" 2>/dev/null || echo "default")
-MODE=$(han parse yaml mode -r --default "OHOTL" < "$INTENT_FILE" 2>/dev/null || echo "OHOTL")
 
 # Resolve workflow to hat sequence
 if [ -f ".ai-dlc/workflows.yml" ]; then
@@ -226,7 +225,7 @@ fi
 FIRST_HAT=$(echo "$WORKFLOW_HATS" | jq -r '.[0]')
 
 # Initialize iteration state
-STATE='{"iteration":1,"hat":"'"${FIRST_HAT}"'","workflowName":"'"${WORKFLOW_NAME}"'","mode":"'"${MODE}"'","workflow":'"${WORKFLOW_HATS}"',"status":"active"}'
+STATE='{"iteration":1,"hat":"'"${FIRST_HAT}"'","workflowName":"'"${WORKFLOW_NAME}"'","workflow":'"${WORKFLOW_HATS}"',"status":"active"}'
 han keep save iteration.json "$STATE"
 ```
 
@@ -360,7 +359,7 @@ Task({
   description: `${FIRST_HAT}: ${unitName}`,
   name: `${FIRST_HAT}-${unitSlug}`,
   team_name: `ai-dlc-${intentSlug}`,
-  mode: modeToAgentTeamsMode(intentMode),
+
   prompt: `
     Execute the ${FIRST_HAT} role for unit ${unitName}.
 
@@ -397,13 +396,6 @@ Task({
 })
 ```
 
-Mode mapping:
-
-| Intent Mode | Task `mode` | Behavior |
-|-------------|-------------|----------|
-| HITL | `plan` | Agent proposes, human approves |
-| OHOTL | `acceptEdits` | Agent works, human can intervene |
-| AHOTL | `bypassPermissions` | Full autonomy within criteria |
 
 ### Step 4 (Teams): Monitor and React Loop
 
@@ -450,7 +442,7 @@ Task({
   description: `${nextHat}: ${unitName}`,
   name: `${nextHat}-${unitSlug}`,
   team_name: `ai-dlc-${intentSlug}`,
-  mode: modeToAgentTeamsMode(intentMode),
+
   prompt: `
     Execute the ${nextHat} role for unit ${unitName}.
 
@@ -585,7 +577,7 @@ Task({
   description: `integrator: ${intentSlug}`,
   name: `integrator-${intentSlug}`,
   team_name: `ai-dlc-${intentSlug}`,
-  mode: modeToAgentTeamsMode(intentMode),
+
   prompt: `
     Execute the Integrator role for intent ${intentSlug}.
 
@@ -800,22 +792,10 @@ STATE=$(echo "$STATE" | han parse json --set "currentUnit=$UNIT_NAME")
 han keep save iteration.json "$STATE"
 ```
 
-#### Step 3: Read Mode and Spawn Subagent
+#### Step 3: Spawn Subagent
 
 **CRITICAL: Do NOT execute hat work inline. Always spawn a subagent.**
 
-##### Read operating mode from intent
-
-```bash
-# Mode is defined on the intent (set during elaboration)
-MODE=$(echo "$STATE" | han parse json mode -r --default OHOTL)
-
-# Also check intent.md frontmatter as source of truth
-if [ -f "${INTENT_DIR}/intent.md" ]; then
-  MODE=$(han parse yaml mode -r --default "$MODE" < "${INTENT_DIR}/intent.md" 2>/dev/null || echo "$MODE")
-fi
-```
-
 ##### Spawn based on `state.hat`
 
 | Role | Agent Type | Description |

plugin/skills/elaborate/SKILL.md

@@ -372,36 +372,6 @@ If the user selects "Custom", ask them to specify which hats to include and in w
 
 ---
 
-## Phase 3.5: Select Operating Mode
-
-The operating mode defines the level of human oversight for this intent. This applies to the **entire intent**, not individual hats.
-
-Use `AskUserQuestion`:
-```json
-{
-  "questions": [{
-    "question": "What level of human oversight for this intent?",
-    "header": "Mode",
-    "options": [
-      {"label": "OHOTL (Recommended)", "description": "AI edits files freely without asking. You see changes in real-time and can interrupt at any point. Best for most development work."},
-      {"label": "HITL", "description": "AI proposes each change as a plan and waits for your approval before editing any file. You approve or reject each step. Best for novel or high-risk work."},
-      {"label": "AHOTL", "description": "AI runs with full autonomy — no permission prompts, no pauses. You review the final result. Best for well-defined, low-risk tasks with clear success criteria."}
-    ],
-    "multiSelect": false
-  }]
-}
-```
-
-**Mode mapping for Agent Teams (when `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS` is enabled):**
-
-| AI-DLC Mode | Agent Teams `mode` | Behavior |
-|-------------|-------------------|----------|
-| HITL | `plan` | Agent proposes, human approves each step |
-| OHOTL | `acceptEdits` | Agent works autonomously, human can intervene |
-| AHOTL | `bypassPermissions` | Agent operates with full autonomy |
-
----
-
 ## Phase 4: Define Success Criteria
 
 Work with the user to define 3-7 **verifiable** success criteria. Each MUST be:
@@ -571,8 +541,8 @@ For each unit:
   - Builds: {specific components/modules/endpoints}
   - Criteria: {count} success criteria
 
-### Workflow & Mode
-{workflow name} — {mode}
+### Workflow
+{workflow name}
 ```
 
 Then ask with `AskUserQuestion`:
@@ -679,7 +649,6 @@ This ensures:
 ```markdown
 ---
 workflow: {workflow-name}
-mode: {HITL|OHOTL|AHOTL}
 git:
   change_strategy: {unit|intent|trunk|bolt}
   auto_merge: {true|false}
@@ -906,7 +875,6 @@ Created: .ai-dlc/{intent-slug}/
 ...
 
 Workflow: {workflowName}
-Mode: {mode}
 Next hat: {next-hat}
 ```
 

plugin/skills/fail/SKILL.md

@@ -89,7 +89,7 @@ Task({
   description: `builder (retry): ${unitName}`,
   name: `builder-${unitSlug}-retry${retries}`,
   team_name: `ai-dlc-${intentSlug}`,
-  mode: modeToAgentTeamsMode(intentMode),
+
   prompt: `
     Re-execute the builder role for unit ${unitName}.