perf: all Claude Code calls use Skills (/fix-cain, /fix-loop)

perf: God 2-step monitoring — diagnose first, Claude Code only when needed

perf: API budget pacing — stretch usage from 3h to 5h

merge: adopt God's 25 iterations on discussion/execution balance

feat: enforce discussion/execution balance — push frequency + trial-and-error

feat: enforce discussion/execution balance — push frequency tracking + trial-and-error

feat: migrate all Claude Code invocations to ACP protocol via acpx

refactor: migrate Adam & Eve from direct Zhipu API to A2A protocol

feat: add agent memory + God silent mode

docs: update README architecture to v3 with God supervisor

feat: inject top priority directive — Cain must use HuggingClaw Docker architecture

feat: add God as supervisor in chat room and conversation loop

fix: add action history to prevent agents from repeating same actions

fix: use correct kebab-case flag --dangerously-skip-permissions

fix: enable Claude Code to actually write files (skip permissions + prompt)

fix: adaptive pacing + anti-repetition when CC output is stale

feat: parallel discussion + async Claude Code execution

fix: strip redundant speaker labels, add timestamps, protect delete_env

feat: distinct father/mother roles + mandatory [TASK] every turn

feat: force every turn to produce a [TASK] for Claude Code

feat: stream Claude Code output to conv-loop log with [CC] prefix

feat: add delete_env action + collision auto-detection

refactor: conversation-loop v2 — Claude Code as primary executor

fix: use --user flag for pip install (non-root container)

fix: install requests at runtime before launching conversation-loop

feat: add /api/conv-loop-log endpoint for remote diagnostics

feat: run conversation-loop on Home Space cloud + CONFIG_ERROR fixes

feat: integrate Claude Code as coding-agent backend (z.ai/Zhipu GLM)

feat: add coding-agent OpenClaw extension with 8 HF Space tools

fix: clear write dedup when Cain is RUNNING but API unresponsive

feat: persist Adam & Eve conversation to HF Dataset

fix: move global declaration to function top to fix SyntaxError

fix: defer cooldown to end of turn so agents can batch file ops

feat: add delete_file action so agents can remove conflicting files

fix: add sdk:gradio vs sdk:docker warning — Dockerfile ignored in gradio mode

fix: add pip uninstall hint for cached deps + torch OOM warning

fix: add HF Docker cache override warning + retry strategy rule

fix: prevent wasted actions from Chinese action names in LLM output

fix: fetch runtime logs in check_health when RUNNING but API unresponsive

fix: add HF Spaces port 7860/Gradio constraint to agent system prompt

fix: allow write_file during APP_STARTING so agents can fix stuck startups

fix: add grace period after smart-wait timeout for consecutive agent turns

fix: add smart-wait timeout to prevent infinite APP_STARTING polling

fix: clear write-dedup and knowledge cache on RUNTIME_ERROR/BUILD_ERROR

fix: change break→continue in action guards to preserve multi-action turns

feat: multi-action turns + sub-agent delegation + GLM-4.5

perf: major efficiency overhaul for conversation-loop mechanism

fix: block write_file during BUILDING/APP_STARTING, fix guard bypass

fix: improve bilingual parsing to split English/Chinese by paragraph

fix: skip forced ACT transition when Cain is BUILDING/RESTARTING