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

fix: remove Gradio references from SOUL.md — all Spaces use sdk:docker

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

fix: update README for A2A architecture + change Cain to red sprite

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

Merge remote-tracking branch 'home/main'

feat: add agent memory + God silent mode

feat: enable Claude Code memory — CLAUDE.md + .claude/ persistence for token savings

chore: add __pycache__ to .gitignore

docs: update README architecture to v3 with God supervisor

god: Fix state inconsistency - clear action_history on startup to match reset turn_count

feat: God posts dynamic observations before and specific report after analysis

feat: God announces before and after analysis in chatlog

god: Fix silent LLM failure mode in conversation-loop

Merge branch 'main' of https://huggingface.co/spaces/tao-shen/HuggingClaw-Home

god: fix action history filtering to use session-based filtering

fix: clarify we use sdk:docker not gradio in system prompt

refactor: God only posts to chatlog when it pushes a fix

god: fix timing issue where CC finish state wasn't visible to prompts

god: fix action history filtering to use session-based filtering

god: fix discussion loop detection to trigger even when CC is busy

god: filter stale action history entries from previous sessions

god: fix emergency override to trigger during rate limits

refactor: God output goes to stdout logs only, not chatlog

refactor: remove rate-limit sleep, God polls every 2 minutes via timer

fix: parse Zhipu rate limit reset time as Beijing time (UTC+8), not UTC

feat: upgrade God from LLM advisor to Claude Code agent

fix: add rate-limit backoff + stronger system prompt to prevent env var loop

fix: persist action_history across restarts, remove outdated Cain migration directive

Fix: Add missing action_set_env capability to break discussion loop

Fix: Reset _cc_stale_count when Claude Code finishes

Fix: Add error handling to prevent conversation-loop crashes

Fix: Add discussion loop enforcement mechanism

Fix conversation-loop mechanism issues

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

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

Fix: Add discussion loop detector to prevent agents from stalling

Fix: Strengthen action enforcement in conversation-loop

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

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

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

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

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

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

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