# Runs deepagent_architect --loop for an EXISTING plan -- processes the
# ENTIRE open backlog from tasks.md automatically, one ticket at a time,
# claiming each via .1bcoder/arch/<plan>/<id>/.claim so multiple concurrent
# --loop processes (different machines/models) can safely divide the work
# without duplicating it. Delegates leaf methods to deepagent_code as it
# goes. Resumable: if killed mid-run, just run this again -- already-done
# tickets are skipped (.done marker), already-decided levels within a
# still-open ticket are skipped too (item_<id>.md already on disk).
#
# Required params:
#   plan = the existing plan name (e.g. plan1)
#
# Usage:
#   1bcoder --host openai://localhost:1234 --model qwen3-4b-instruct-2507 ^
#       --scriptapply deepagent_loop.txt --param plan=plan1
/param timeout 300
/flow deepagent_architect {{plan}} --loop --lang py
