Execute implementation plans in background while Pichu stays responsive. Wraps subagent-driven-development with run_in_background: true.
Execute plans in background while staying responsive to new messages.
After writing-plans skill completes and user approves execution.
User can also trigger manually: "execute the plan"
./scripts/log-session.sh {chat_id} task_started {task_id} "{description}"Task(tool) - see references/task-template.md./scripts/log-session.sh {chat_id} task_completed {task_id} "{summary}"run_in_background: truescripts/task-state.shAfter subagent completes, capture the observation:
# After TaskOutput is received
TASK_ID="<task_id>" \
TASK_TYPE="<implement|fix|review>" \
TASK_PROMPT="<original_prompt>" \
TASK_RESULT="<result_summary>" \
FILES_CHANGED="<file1 file2>" \
./scripts/hooks/after-complete.sh
The hook appends to state/learning/observations.jsonl for later processing by /evolve.
writing-plans - creates the plansubagent-driven-development - execution enginecommander - orchestrates flow and notificationsIf task fails:
Escalation Triggers:
Escalation Message: "Task failed after 3 attempts: [error summary]. Options: Modify plan, Cancel, Try different approach?"