Continue an epic with interactive checkpoints
Epic name: Provided via arguments
Phases: Spec → Research → Plan → Execute
Check .state for yolo status:
If yolo.active: true:
Note: Yolo mode was active. Switching to interactive mode.
Update .state:
{
"yolo": {
"active": false
}
}
If no spec:
No spec found for epic '<name>'.
Create the epic first: /lisa-create-epic <name>
STOP.
If spec exists: Read and briefly summarize, then continue to step 2.
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:
ATTEMPT 1:
build_research_context("${epicName}") to get the research promptverify_research_complete("${epicName}")ATTEMPT 2 (if attempt 1 incomplete):
[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]
verify_research_complete("${epicName}") againATTEMPT 3 (if attempt 2 incomplete):
[FINAL ATTEMPT - COMPLETE OR FAIL]
This is your FINAL attempt. If incomplete, task will be marked BLOCKED.
STILL MISSING: [list remaining issues]
DO EXACTLY THIS:
1. Read spec.md
2. Write research.md with ALL required sections
3. Update .state file
4. Confirm completion
NO EXCUSES. DO IT NOW.
verify_research_complete("${epicName}").state to set researchComplete: false with note about issuesIf 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:
Same retry pattern as Research (3 attempts max):
ATTEMPT 1:
build_planning_context("${epicName}") to get the planning promptverify_planning_complete("${epicName}")ATTEMPT 2-3: Same retry logic with stronger prompts
If still incomplete after 3 attempts:
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:
Use config execution.maxParallelTasks (default: 1).
For each task:
Call get_available_tasks(epicName)
Take up to maxParallelTasks from available list
For each task:
build_task_context(epicName, taskId)verify_task_complete(epicName, taskId)If verification.complete === true: Continue to next task
If verification.complete === false:
RETRY ATTEMPT 2:
[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]
verify_task_complete(epicName, taskId) againRETRY ATTEMPT 3:
IF STILL INCOMPLETE:
blocked in task fileAfter all available tasks processed, call get_available_tasks again
Show progress, ask "Continue with next batch?"
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]"