Lisa Continue

STOP.

If spec exists: Read and briefly summarize, then continue to step 2.

Step 2: Research Phase

If research.md exists:

"Research already complete. Proceeding to planning..." Skip to step 3.

If not done:

"Ready to start research? I'll explore the codebase with a dedicated agent."

Wait for confirmation. On yes:

Research Execution with Verification

ATTEMPT 1:

1. Call build_research_context("${epicName}") to get the research prompt
2. Call Task tool with:
   • prompt: the returned prompt from build_research_context
   • subagent_type: "explore"
3. After Task completes, call verify_research_complete("${epicName}")
4. If verification.complete === true:
   • Show: "Research complete. Continue to planning?"
   • STOP and wait for user confirmation

ATTEMPT 2 (if attempt 1 incomplete):

1. Call Task tool again with STRONGER prompt:

   [RESEARCH RETRY - MUST COMPLETE]
   
   Your previous attempt did not complete successfully.
   MISSING: [list issues from verification.issues]
   
   MANDATORY ACTIONS - DO NOT STOP UNTIL DONE:
   1. Read spec.md to understand requirements
   2. Explore codebase thoroughly
   3. Save COMPLETE research.md with ALL sections
   4. Update .state with researchComplete: true
   
   You MUST complete all items above. DO NOT stop early.
   DO NOT say "I will do this" - DO IT NOW.
   
   [original prompt from build_research_context]
   

2. After Task completes, call verify_research_complete("${epicName}") again
3. If verification.complete === true: Ask to continue to planning

ATTEMPT 3 (if attempt 2 incomplete):

1. Call Task tool one more time with FINAL prompt:

   [FINAL ATTEMPT - COMPLETE OR FAIL]
   
   This is your FINAL attempt. If incomplete, task will be marked BLOCKED.
   
   STILL MISSING: [list remaining issues]
   
   DO EXACTLY THIS:
   1. Read spec.md
   2. Write research.md with ALL required sections
   3. Update .state file
   4. Confirm completion
   
   NO EXCUSES. DO IT NOW.
   

2. After Task completes, call verify_research_complete("${epicName}")
3. If still incomplete:
   • Update .state to set researchComplete: false with note about issues
   • Show: "Research could not be completed after 3 attempts. Issues: [list]. You can retry with /lisa-continue ${epicName} or mark research as manually complete."
   • STOP

Step 3: Plan Phase

If plan.md exists:

"Plan already complete with X tasks. Proceeding to execution..." Skip to step 4.

If not done:

"Ready to create the implementation plan?"

Wait for confirmation. On yes:

Planning Execution with Verification

Same retry pattern as Research (3 attempts max):

ATTEMPT 1:

1. Call build_planning_context("${epicName}") to get the planning prompt
2. Call Task tool with:
   • prompt: the returned prompt from build_planning_context
   • subagent_type: "general"
3. After Task completes, call verify_planning_complete("${epicName}")
4. If verification.complete === true: Show summary and ask to continue to execution

ATTEMPT 2-3: Same retry logic with stronger prompts

If still incomplete after 3 attempts:

• Show: "Planning could not be completed after 3 attempts. Issues: [list]. You can retry or manually create the plan."
• STOP

Step 4: Execute Phase

Use get_available_tasks tool to check status.

If all tasks done:

"All tasks complete! Epic finished." STOP.

If tasks remain: Show summary and ask:

"Ready to execute? X tasks remaining:

• Available now: [list]
• Blocked: [list]"

Wait for confirmation. On yes:

Execution Loop with Verification

Use config execution.maxParallelTasks (default: 1).

For each task:

1. Call get_available_tasks(epicName)

2. Take up to maxParallelTasks from available list

3. For each task:

   • Call build_task_context(epicName, taskId)
   • Call Task tool with the prompt
   • After Task completes, call verify_task_complete(epicName, taskId)

4. If verification.complete === true: Continue to next task

5. If verification.complete === false:

   RETRY ATTEMPT 2:

   • Call Task tool with STRONGER prompt:

     [TASK RETRY - MUST COMPLETE]
     
     Task ${taskId} was NOT completed properly.
     MISSING: [list verification.issues]
     
     You MUST:
     1. Complete the task described in ${taskFile}
     2. Update status to "done"
     3. Add ## Report section with ### What Was Done and ### Files Changed
     4. DO NOT stop until ALL criteria are met
     
     DO IT NOW. NO EXCUSES.
     
     [original task prompt]
     

   • After Task completes, call verify_task_complete(epicName, taskId) again

   RETRY ATTEMPT 3:

   • If still incomplete, final attempt with URGENT prompt
   • After Task completes, call verification

   IF STILL INCOMPLETE:

   • Mark task as blocked in task file
   • Show: "Task ${taskId} blocked after 3 attempts. Moving to next task."
   • Continue with other tasks

6. After all available tasks processed, call get_available_tasks again

7. Show progress, ask "Continue with next batch?"

8. Repeat until all done or all remaining tasks blocked

On all complete: Check config git.completionMode and handle accordingly (see COMMON.md).

Show:

"Epic complete! All X tasks finished.

Changes made: [summary]"