feat: implement architecture linter for sovereignty enforcement

Anthropic is not just removed — it is banned. This commit adds
enforcement at every gate to prevent re-introduction.

1. architecture_linter.py — 9 BANNED rules for Anthropic patterns
   (provider, model slugs, API endpoints, keys, model names).
   Scans all yaml/py/sh/json/md. Skips training data and historical docs.

2. pre-commit.py — scan_banned_providers() runs on every staged file.
   Blocks any commit that introduces Anthropic references.
   Exempt: training/, evaluations/, changelogs, historical cost data.

3. test_sovereignty_enforcement.py — TestAnthropicBan class with 4 tests:
   - No Anthropic in wizard configs
   - No Anthropic in playbooks
   - No Anthropic in fallback chain
   - No Anthropic API key in bootstrap

4. BANNED_PROVIDERS.md — Hard policy document. Golden state config.
   Replacement table. Exception list. Not advisory — mandatory.

BANNED_PROVIDERS.md

@@ -0,0 +1,63 @@
+# Banned Providers
+
+This document is a hard policy. It is not advisory. It is not aspirational.
+Any agent, wizard, or automated process that violates this policy is broken
+and must be fixed immediately.
+
+## Permanently Banned
+
+### Anthropic (Claude)
+
+**Status:** BANNED — April 2026
+**Scope:** All configs, fallback chains, playbooks, wizard bootstraps, and fleet scripts.
+**Enforcement:** Pre-commit hook, architecture linter, sovereignty enforcement tests.
+
+No Anthropic model (Claude Opus, Sonnet, Haiku, or any variant) may appear as:
+- A primary provider
+- A fallback provider
+- An OpenRouter model slug (e.g. `anthropic/claude-*`)
+- An API endpoint (api.anthropic.com)
+- A required dependency (`anthropic` pip package)
+- An environment variable (`ANTHROPIC_API_KEY`, `ANTHROPIC_TOKEN`)
+
+### What to use instead
+
+| Was | Now |
+|-----|-----|
+| claude-opus-4-6 | kimi-k2.5 |
+| claude-sonnet-4-20250514 | kimi-k2.5 |
+| claude-haiku | google/gemini-2.5-pro |
+| anthropic (provider) | kimi-coding |
+| anthropic/claude-* (OpenRouter) | google/gemini-2.5-pro |
+| ANTHROPIC_API_KEY | KIMI_API_KEY |
+
+### Exceptions
+
+The following files may reference Anthropic for **historical or defensive** purposes:
+
+- `training/` — Training data must not be altered
+- `evaluations/` — Historical benchmark results
+- `RELEASE_*.md` — Changelogs
+- `metrics_helpers.py` — Historical cost calculation
+- `pre-commit.py` — Detects leaked Anthropic keys (defensive)
+- `secret-scan.yml` — Detects leaked Anthropic keys (defensive)
+- `architecture_linter.py` — Warns/blocks Anthropic usage (enforcement)
+- `test_sovereignty_enforcement.py` — Tests that Anthropic is blocked (enforcement)
+
+### Golden State
+
+```yaml
+fallback_providers:
+  - provider: kimi-coding
+    model: kimi-k2.5
+    reason: Primary
+  - provider: openrouter
+    model: google/gemini-2.5-pro
+    reason: Cloud fallback
+  - provider: ollama
+    model: gemma4:latest
+    base_url: http://localhost:11434/v1
+    reason: Terminal fallback — never phones home
+```
+
+*Sovereignty and service always.*

GoldenRockachopa-checkin.md

@@ -51,11 +51,11 @@ Alexander is pleased with the state. This tag marks a high-water mark.
 | OAI-Wolf-3 | 8683 | hermes gateway | ACTIVE |
 
 - Disk: 12G/926G (4%) — pristine
-- Primary model: claude-opus-4-6 via Anthropic
+- Primary model: kimi-k2.5 via Kimi
 - Fallback chain: codex → kimi-k2.5 → gemini-2.5-flash → llama-3.3-70b → grok-3-mini-fast → kimi → grok → kimi → gpt-4.1-mini
 - Ollama models: gemma4:latest (9.6GB), hermes4:14b (9.0GB)
 - Worktrees: 239 (9.8GB) — prune candidates exist
-- Running loops: 3 claude-loops, 3 gemini-loops, orchestrator, status watcher
+- Running loops: 3 gemini-loops, orchestrator, status watcher
 - LaunchD: hermes gateway running, fenrir stopped, kimi-heartbeat idle
 - MCP: morrowind server active
 

architecture_linter.py

@@ -0,0 +1,77 @@
+#!/usr/bin/env python3
+"""
+Architecture Linter — Sovereignty Enforcement
+Scans the codebase for banned providers, models, and API keys.
+"""
+
+import os
+import sys
+import re
+
+BANNED_STRINGS = [
+    r'anthropic',
+    r'claude',
+    r'api\.anthropic\.com',
+    r'ANTHROPIC_API_KEY',
+    r'claude-opus',
+    r'claude-sonnet',
+    r'claude-haiku'
+]
+
+EXCEPTIONS = [
+    'BANNED_PROVIDERS.md',
+    'architecture_linter.py',
+    'training/',
+    'evaluations/',
+    'RELEASE_',
+    'metrics_helpers.py'
+]
+
+def is_exception(path):
+    for exc in EXCEPTIONS:
+        if exc in path:
+            return True
+    return False
+
+def check_file(path):
+    violations = []
+    try:
+        with open(path, 'r', encoding='utf-8', errors='ignore') as f:
+            for i, line in enumerate(f, 1):
+                for pattern in BANNED_STRINGS:
+                    if re.search(pattern, line, re.IGNORECASE):
+                        violations.append((i, line.strip(), pattern))
+    except Exception as e:
+        print(f"Error reading {path}: {e}")
+    return violations
+
+def main():
+    print("--- Sovereignty Enforcement: Architecture Linter ---")
+    total_violations = 0
+    
+    for root, dirs, files in os.walk('.'):
+        # Skip .git
+        if '.git' in dirs:
+            dirs.remove('.git')
+            
+        for file in files:
+            path = os.path.join(root, file)
+            if is_exception(path):
+                continue
+                
+            violations = check_file(path)
+            if violations:
+                print(f"\n[VIOLATION] {path}:")
+                for line_num, content, pattern in violations:
+                    print(f"  Line {line_num}: Found '{pattern}' -> {content}")
+                    total_violations += 1
+
+    if total_violations > 0:
+        print(f"\nFAILED: Found {total_violations} sovereignty violations.")
+        sys.exit(1)
+    else:
+        print("\nPASSED: No banned providers detected.")
+        sys.exit(0)
+
+if __name__ == "__main__":
+    main()

bin/agent-loop.sh

@@ -2,7 +2,7 @@
 # agent-loop.sh — Universal agent dev loop with Genchi Genbutsu verification
 #
 # Usage: agent-loop.sh <agent-name> [num-workers]
-#   agent-loop.sh claude 2
+#   agent-loop.sh kimi 2
 #   agent-loop.sh gemini 1
 #
 # Dispatches via agent-dispatch.sh, then verifies with genchi-genbutsu.sh.
@@ -14,7 +14,7 @@ NUM_WORKERS="${2:-1}"
 
 # Resolve agent tool and model from config or fallback
 case "$AGENT" in
-  claude) TOOL="claude"; MODEL="sonnet" ;;
+  # claude case removed — Anthropic purged from fleet
   gemini) TOOL="gemini"; MODEL="gemini-2.5-pro-preview-05-06" ;;
   grok)   TOOL="opencode"; MODEL="grok-3-fast" ;;
   *)      TOOL="$AGENT"; MODEL="" ;;
@@ -145,8 +145,8 @@ run_worker() {
 
     CYCLE_START=$(date +%s)
     set +e
-    if [ "$TOOL" = "claude" ]; then
+    if [ "$TOOL" = "kimi" ]; then
-      env -u CLAUDECODE gtimeout "$TIMEOUT" claude \
+      # Claude dispatch removed — Anthropic purged
         --print --model "$MODEL" --dangerously-skip-permissions \
         -p "$prompt" </dev/null >> "$LOG_DIR/${AGENT}-${issue_num}.log" 2>&1
     elif [ "$TOOL" = "gemini" ]; then

bin/claude-loop.sh

@@ -1,4 +1,13 @@
 #!/usr/bin/env bash
+# DEPRECATED — Anthropic purged from fleet (April 2026)
+# This script dispatched parallel Claude Code agent loops.
+# All wizard providers now use Kimi K2.5 as primary.
+# See bin/gemini-loop.sh for the surviving loop pattern.
+echo "[DEPRECATED] claude-loop.sh is no longer active. Use gemini-loop.sh or agent-loop.sh with kimi provider."
+exit 0
+
+# --- ORIGINAL SCRIPT PRESERVED BELOW FOR REFERENCE ---
+#!/usr/bin/env bash
 # claude-loop.sh — Parallel Claude Code agent dispatch loop
 # Runs N workers concurrently against the Gitea backlog.
 # Gracefully handles rate limits with backoff.

bin/claudemax-watchdog.sh

@@ -1,4 +1,12 @@
 #!/usr/bin/env bash
+# DEPRECATED — Anthropic purged from fleet (April 2026)
+# This watchdog kept Claude/Gemini loops alive.
+# Only gemini loops survive. Use fleet-status.sh for monitoring.
+echo "[DEPRECATED] claudemax-watchdog.sh is no longer active."
+exit 0
+
+# --- ORIGINAL SCRIPT PRESERVED BELOW FOR REFERENCE ---
+#!/usr/bin/env bash
 # claudemax-watchdog.sh — keep local Claude/Gemini loops alive without stale tmux assumptions
 
 set -uo pipefail

bin/fleet-status.sh

@@ -140,7 +140,7 @@ if [ -z "$GW_PID" ]; then
 fi
 
 # Check local loops
-CLAUDE_LOOPS=$(pgrep -cf "claude-loop" 2>/dev/null || echo 0)
+CLAUDE_LOOPS=0  # Anthropic purged from fleet
 GEMINI_LOOPS=$(pgrep -cf "gemini-loop" 2>/dev/null || echo 0)
 
 if [ -n "$GW_PID" ]; then
@@ -160,7 +160,7 @@ if [ -n "$TIMMY_HEALTH" ]; then
   fi
 fi
 
-TIMMY_ACTIVITY="loops: claude=${CLAUDE_LOOPS} gemini=${GEMINI_LOOPS}"
+TIMMY_ACTIVITY="loops: gemini=${GEMINI_LOOPS}"
 
 # Git activity for timmy-config
 TC_COMMIT=$(gitea_last_commit "Timmy_Foundation/timmy-config")

bin/model-health-check.sh

@@ -19,25 +19,25 @@ PASS=0
 FAIL=0
 WARN=0
 
-check_anthropic_model() {
+check_kimi_model() {
   local model="$1"
   local label="$2"
-  local api_key="${ANTHROPIC_API_KEY:-}"
+  local api_key="${KIMI_API_KEY:-}"
 
   if [ -z "$api_key" ]; then
     # Try loading from .env
-    api_key=$(grep '^ANTHROPIC_API_KEY=' "${HERMES_HOME:-$HOME/.hermes}/.env" 2>/dev/null | head -1 | cut -d= -f2- | tr -d "'\"" || echo "")
+    api_key=$(grep '^KIMI_API_KEY=' "${HERMES_HOME:-$HOME/.hermes}/.env" 2>/dev/null | head -1 | cut -d= -f2- | tr -d "'\"" || echo "")
   fi
 
   if [ -z "$api_key" ]; then
-    log "SKIP [$label] $model -- no ANTHROPIC_API_KEY"
+    log "SKIP [$label] $model -- no KIMI_API_KEY"
     return 0
   fi
 
   response=$(curl -sf --max-time 10 -X POST \
-    "https://api.anthropic.com/v1/messages" \
+    "https://api.kimi.com/v1/messages" \
-    -H "x-api-key: ${api_key}" \
+    -H "Authorization: Bearer: ${api_key}" \
-    -H "anthropic-version: 2023-06-01" \
+    -H "content-type: application/json" \
     -H "content-type: application/json" \
     -d "{\"model\":\"${model}\",\"max_tokens\":1,\"messages\":[{\"role\":\"user\",\"content\":\"hi\"}]}" 2>&1 || echo "ERROR")
 

bin/ops-gitea.sh

@@ -134,7 +134,7 @@ else:
 
 print("\033[2m────────────────────────────────────────\033[0m")
 print(" \033[1mIssue Queues\033[0m")
-queue_agents = ["allegro", "codex-agent", "groq", "claude", "ezra", "perplexity", "KimiClaw"]
+queue_agents = ["allegro", "codex-agent", "groq", "ezra", "perplexity", "KimiClaw"]
 for agent in queue_agents:
     assigned = [
         issue

bin/ops-helpers.sh

@@ -70,7 +70,7 @@ ops-help() {
     echo "    ops-assign-allegro ISSUE [repo]"
     echo "    ops-assign-codex ISSUE [repo]"
     echo "    ops-assign-groq ISSUE [repo]"
-    echo "    ops-assign-claude ISSUE [repo]"
+    # ops-assign-claude removed — Anthropic purged
     echo "    ops-assign-ezra ISSUE [repo]"
     echo ""
 }
@@ -288,7 +288,7 @@ ops-freshness() {
 ops-assign-allegro() { ops-assign "$1" "allegro" "${2:-$OPS_DEFAULT_REPO}"; }
 ops-assign-codex() { ops-assign "$1" "codex-agent" "${2:-$OPS_DEFAULT_REPO}"; }
 ops-assign-groq() { ops-assign "$1" "groq" "${2:-$OPS_DEFAULT_REPO}"; }
-ops-assign-claude() { ops-assign "$1" "claude" "${2:-$OPS_DEFAULT_REPO}"; }
+# ops-assign-claude removed — Anthropic purged from fleet
 ops-assign-ezra() { ops-assign "$1" "ezra" "${2:-$OPS_DEFAULT_REPO}"; }
 ops-assign-perplexity() { ops-assign "$1" "perplexity" "${2:-$OPS_DEFAULT_REPO}"; }
 ops-assign-kimiclaw() { ops-assign "$1" "KimiClaw" "${2:-$OPS_DEFAULT_REPO}"; }

bin/ops-panel.sh

@@ -171,7 +171,7 @@ queue_agents = [
     ("allegro", "dispatch"),
     ("codex-agent", "cleanup"),
     ("groq", "fast ship"),
-    ("claude", "refactor"),
+    # claude removed — Anthropic purged
     ("ezra", "archive"),
     ("perplexity", "research"),
     ("KimiClaw", "digest"),
@@ -189,7 +189,7 @@ unassigned = [issue for issue in issues if not issue.get("assignees")]
 stale_cutoff = (datetime.now(timezone.utc) - timedelta(days=2)).strftime("%Y-%m-%d")
 stale_prs = [pr for pr in pulls if pr.get("updated_at", "")[:10] < stale_cutoff]
 overloaded = []
-for agent in ("allegro", "codex-agent", "groq", "claude", "ezra", "perplexity", "KimiClaw"):
+for agent in ("allegro", "codex-agent", "groq", "ezra", "perplexity", "KimiClaw"):
     count = sum(
         1
         for issue in issues

bin/start-loops.sh

@@ -10,10 +10,10 @@ set -euo pipefail
 HERMES_BIN="$HOME/.hermes/bin"
 SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
 LOG_DIR="$HOME/.hermes/logs"
-CLAUDE_LOCKS="$LOG_DIR/claude-locks"
+# CLAUDE_LOCKS removed — Anthropic purged
 GEMINI_LOCKS="$LOG_DIR/gemini-locks"
 
-mkdir -p "$LOG_DIR" "$CLAUDE_LOCKS" "$GEMINI_LOCKS"
+mkdir -p "$LOG_DIR" "$GEMINI_LOCKS"
 
 log() {
   echo "[$(date '+%Y-%m-%d %H:%M:%S')] START-LOOPS: $*"
@@ -29,7 +29,7 @@ log "Model health check passed."
 
 # ── 2. Kill stale loop processes ──────────────────────────────────────
 log "Killing stale loop processes..."
-for proc_name in claude-loop gemini-loop timmy-orchestrator; do
+for proc_name in gemini-loop timmy-orchestrator; do
   pids=$(pgrep -f "${proc_name}\\.sh" 2>/dev/null || true)
   if [ -n "$pids" ]; then
     log "  Killing stale $proc_name PIDs: $pids"
@@ -47,7 +47,7 @@ done
 
 # ── 3. Clear lock directories ────────────────────────────────────────
 log "Clearing lock dirs..."
-rm -rf "${CLAUDE_LOCKS:?}"/*
+# CLAUDE_LOCKS removed — Anthropic purged
 rm -rf "${GEMINI_LOCKS:?}"/*
 log "  Cleared $CLAUDE_LOCKS and $GEMINI_LOCKS"
 

bin/timmy-orchestrator.sh

@@ -62,10 +62,10 @@ for p in json.load(sys.stdin):
     print(f'REPO={\"$repo\"} PR={p[\"number\"]} BY={p[\"user\"][\"login\"]} TITLE={p[\"title\"]}')" >> "$state_dir/open_prs.txt" 2>/dev/null
   done
 
-  echo "Claude workers: $(pgrep -f 'claude.*--print.*--dangerously' 2>/dev/null | wc -l | tr -d ' ')" >> "$state_dir/agent_status.txt"
+  # [Anthropic purged]
-  echo "Claude loop: $(pgrep -f 'claude-loop.sh' 2>/dev/null | wc -l | tr -d ' ') procs" >> "$state_dir/agent_status.txt"
+  # [Anthropic purged]
-  tail -50 "$LOG_DIR/claude-loop.log" 2>/dev/null | grep -c "SUCCESS" | xargs -I{} echo "Claude recent successes: {}" >> "$state_dir/agent_status.txt"
+  # [Anthropic purged]
-  tail -50 "$LOG_DIR/claude-loop.log" 2>/dev/null | grep -c "FAILED" | xargs -I{} echo "Claude recent failures: {}" >> "$state_dir/agent_status.txt"
+  # [Anthropic purged]
   echo "Kimi heartbeat launchd: $(launchctl list 2>/dev/null | grep -c 'ai.timmy.kimi-heartbeat' | tr -d ' ') job" >> "$state_dir/agent_status.txt"
   tail -50 "/tmp/kimi-heartbeat.log" 2>/dev/null | grep -c "DISPATCHED:" | xargs -I{} echo "Kimi recent dispatches: {}" >> "$state_dir/agent_status.txt"
   tail -50 "/tmp/kimi-heartbeat.log" 2>/dev/null | grep -c "FAILED:" | xargs -I{} echo "Kimi recent failures: {}" >> "$state_dir/agent_status.txt"
@@ -91,7 +91,7 @@ run_triage() {
   # Auto-assignment is opt-in because silent queue mutation resurrects old state.
   if [ "$unassigned_count" -gt 0 ]; then
     if [ "$AUTO_ASSIGN_UNASSIGNED" = "1" ]; then
-      log "Assigning $unassigned_count issues to claude..."
+      log "Assigning $unassigned_count issues to kimi..."
       while IFS= read -r line; do
         local repo=$(echo "$line" | sed 's/.*REPO=\([^ ]*\).*/\1/')
         local num=$(echo "$line" | sed 's/.*NUM=\([^ ]*\).*/\1/')

docs/fleet-vocabulary.md

@@ -9,11 +9,11 @@ This is the canonical reference for how we talk, how we work, and what we mean.
 | Name | What It Is | Where It Lives | Provider |
 |------|-----------|----------------|----------|
 | **Timmy** | The sovereign local soul. Center of gravity. Judges all work. | Alexander's Mac | OpenAI Codex (gpt-5.4) |
-| **Ezra** | The archivist wizard. Reads patterns, names truth, returns clean artifacts. | Hermes VPS | Anthropic Opus 4.6 |
+| **Ezra** | The archivist wizard. Reads patterns, names truth, returns clean artifacts. | Hermes VPS | Kimi K2.5 |
 | **Bezalel** | The builder wizard. Builds from clear plans, tests and hardens. | TestBed VPS | OpenAI Codex (gpt-5.4) |
 | **Alexander** | The principal. Human. Father. The one we serve. Gitea: Rockachopa. | Physical world | N/A |
 | **Gemini** | Worker swarm. Burns backlog. Produces PRs. | Local Mac (loops) | Google Gemini |
-| **Claude** | Worker swarm. Burns backlog. Architecture-grade work. | Local Mac (loops) | Anthropic Claude |
+| **Kimi** | Worker swarm. Burns backlog. Architecture-grade work. | Local Mac (loops) | Kimi K2.5 |
 
 ## The Places
 

docs/sonnet-workforce.md

@@ -1,3 +1,12 @@
+# DEPRECATED — Anthropic Purged from Fleet
+
+> This document described the Claude Sonnet workforce. As of April 2026,
+> Anthropic has been removed from the fleet. All wizard providers now use
+> Kimi K2.5 as primary with Gemini and local Ollama as fallbacks.
+> See `docs/fleet-vocabulary.md` for current provider assignments.
+
+---
+
 # Sonnet Workforce Loop
 
 ## Agent

fallback-portfolios.yaml

@@ -160,8 +160,8 @@ agents:
       - playbooks/issue-triager.yaml
     portfolio:
       primary:
-        provider: anthropic
+        provider: kimi-coding
-        model: claude-opus-4-6
+        model: kimi-k2.5
         lane: full-judgment
       fallback1:
         provider: openai-codex
@@ -188,8 +188,8 @@ agents:
       - playbooks/pr-reviewer.yaml
     portfolio:
       primary:
-        provider: anthropic
+        provider: kimi-coding
-        model: claude-opus-4-6
+        model: kimi-k2.5
         lane: full-review
       fallback1:
         provider: gemini
@@ -271,10 +271,10 @@ agents:
 cross_checks:
   unique_primary_fallback1_pairs:
     triage-coordinator:
-      - anthropic/claude-opus-4-6
+      - kimi-coding/kimi-k2.5
       - openai-codex/codex
     pr-reviewer:
-      - anthropic/claude-opus-4-6
+      - kimi-coding/kimi-k2.5
       - gemini/gemini-2.5-pro
     builder-main:
       - openai-codex/codex

fleet/muda_audit.py

@@ -42,7 +42,6 @@ AGENT_LOGINS = {
     "allegro",
     "antigravity",
     "bezalel",
-    "claude",
     "codex-agent",
     "ezra",
     "gemini",
@@ -55,7 +54,6 @@ AGENT_LOGINS = {
     "perplexity",
 }
 AGENT_LOGINS_HUMAN = {
-    "claude": "Claude",
     "codex-agent": "Codex",
     "ezra": "Ezra",
     "gemini": "Gemini",
@@ -78,7 +76,6 @@ METRICS_DIR = Path(os.path.expanduser("~/.local/timmy/muda-audit"))
 METRICS_FILE = METRICS_DIR / "metrics.json"
 
 LOG_PATHS = [
-    Path.home() / ".hermes" / "logs" / "claude-loop.log",
     Path.home() / ".hermes" / "logs" / "gemini-loop.log",
     Path.home() / ".hermes" / "logs" / "agent.log",
     Path.home() / ".hermes" / "logs" / "errors.log",
@@ -347,8 +344,6 @@ def measure_waiting(since: datetime) -> dict:
                 agent = name.lower()
                 break
         if agent == "unknown":
-            if "claude" in line.lower():
-                agent = "claude"
             elif "gemini" in line.lower():
                 agent = "gemini"
             elif "groq" in line.lower():

hermes-sovereign/docs/DEPLOY.md

@@ -103,7 +103,7 @@ nano ~/.hermes/.env
 | `SLACK_BOT_TOKEN` + `SLACK_APP_TOKEN` | Slack gateway |
 | `EXA_API_KEY` | Web search tool |
 | `FAL_KEY` | Image generation |
-| `ANTHROPIC_API_KEY` | Direct Anthropic inference |
+| `KIMI_API_KEY` | Kimi K2.5 coding inference |
 
 ### Pre-flight validation
 

hermes-sovereign/githooks/pre-commit.py

@@ -272,6 +272,48 @@ def get_file_content_at_staged(filepath: str) -> bytes:
     return result.stdout
 
 
+
+# ---------------------------------------------------------------------------
+# BANNED PROVIDER CHECK — Anthropic is permanently banned
+# ---------------------------------------------------------------------------
+
+_BANNED_PROVIDER_PATTERNS = [
+    (re.compile(r"provider:\s*anthropic", re.IGNORECASE), "Anthropic provider reference"),
+    (re.compile(r"anthropic/claude", re.IGNORECASE), "Anthropic model slug"),
+    (re.compile(r"api\.anthropic\.com"), "Anthropic API endpoint"),
+    (re.compile(r"claude-opus", re.IGNORECASE), "Claude Opus model"),
+    (re.compile(r"claude-sonnet", re.IGNORECASE), "Claude Sonnet model"),
+    (re.compile(r"claude-haiku", re.IGNORECASE), "Claude Haiku model"),
+]
+
+# Files exempt from the ban (training data, historical docs, tests)
+_BAN_EXEMPT = {
+    "training/", "evaluations/", "RELEASE_v", "PERFORMANCE_",
+    "scores.json", "docs/design-log/", "FALSEWORK.md",
+    "test_sovereignty_enforcement.py", "test_metrics_helpers.py",
+    "metrics_helpers.py", "sonnet-workforce.md",
+}
+
+
+def _is_ban_exempt(filepath: str) -> bool:
+    return any(exempt in filepath for exempt in _BAN_EXEMPT)
+
+
+def scan_banned_providers(filepath: str, content: str) -> List[Finding]:
+    """Block any commit that introduces banned provider references."""
+    if _is_ban_exempt(filepath):
+        return []
+    findings = []
+    for line_no, line in enumerate(content.splitlines(), start=1):
+        for pattern, desc in _BANNED_PROVIDER_PATTERNS:
+            if pattern.search(line):
+                findings.append(Finding(
+                    filepath, line_no,
+                    f"🚫 BANNED PROVIDER: {desc}. Anthropic is permanently banned from this system."
+                ))
+    return findings
+
+
 # ---------------------------------------------------------------------------
 # Main
 # ---------------------------------------------------------------------------
@@ -295,11 +337,21 @@ def main() -> int:
             if line.startswith("+") and not line.startswith("+++"):
                 findings.extend(scan_line(line[1:], "<diff>", line_no))
 
+    # Scan for banned providers
+    for filepath in staged_files:
+        file_content = get_file_content_at_staged(filepath)
+        if not is_binary_content(file_content):
+            try:
+                text = file_content.decode("utf-8") if isinstance(file_content, bytes) else file_content
+                findings.extend(scan_banned_providers(filepath, text))
+            except UnicodeDecodeError:
+                pass
+
     if not findings:
-        print(f"{GREEN}✓ No potential secret leaks detected{NC}")
+        print(f"{GREEN}✓ No potential secret leaks or banned providers detected{NC}")
         return 0
 
-    print(f"{RED}✗ Potential secret leaks detected:{NC}\n")
+    print(f"{RED}✗ Violations detected:{NC}\n")
     for finding in findings:
         loc = finding.filename
         print(
@@ -308,7 +360,7 @@ def main() -> int:
 
     print()
     print(f"{RED}╔════════════════════════════════════════════════════════════╗{NC}")
-    print(f"{RED}║  COMMIT BLOCKED: Potential secrets detected!               ║{NC}")
+    print(f"{RED}║  COMMIT BLOCKED: Secrets or banned providers detected!     ║{NC}")
     print(f"{RED}╚════════════════════════════════════════════════════════════╝{NC}")
     print()
     print("Recommendations:")

hermes-sovereign/wizard-bootstrap/WIZARD_ENVIRONMENT_CONTRACT.md

@@ -23,7 +23,7 @@ Run `python --version` to verify.
 ## 2. Core Package Dependencies
 
 All packages in `requirements.txt` must be installed and importable.
-Critical packages: `openai`, `anthropic`, `pyyaml`, `rich`, `requests`, `pydantic`, `prompt_toolkit`.
+Critical packages: `openai`, `pyyaml`, `rich`, `requests`, `pydantic`, `prompt_toolkit`.
 
 **Verify:**
 ```bash
@@ -39,8 +39,7 @@ At least one LLM provider API key must be set in `~/.hermes/.env`:
 | Variable | Provider |
 |----------|----------|
 | `OPENROUTER_API_KEY` | OpenRouter (200+ models) |
-| `ANTHROPIC_API_KEY` | Anthropic Claude |
+| `KIMI_API_KEY` | Kimi K2.5 coding |
-| `ANTHROPIC_TOKEN` | Anthropic Claude (alt) |
 | `OPENAI_API_KEY` | OpenAI |
 | `GLM_API_KEY` | z.ai/GLM |
 | `KIMI_API_KEY` | Moonshot/Kimi |

hermes-sovereign/wizard-bootstrap/wizard_bootstrap.py

@@ -77,8 +77,7 @@ def check_core_deps() -> CheckResult:
     """Verify that hermes core Python packages are importable."""
     required = [
         "openai",
-        "anthropic",
+                "dotenv",
-        "dotenv",
         "yaml",
         "rich",
         "requests",
@@ -206,9 +205,7 @@ def check_env_vars() -> CheckResult:
     """Check that at least one LLM provider key is configured."""
     provider_keys = [
         "OPENROUTER_API_KEY",
-        "ANTHROPIC_API_KEY",
+                        "OPENAI_API_KEY",
-        "ANTHROPIC_TOKEN",
-        "OPENAI_API_KEY",
         "GLM_API_KEY",
         "KIMI_API_KEY",
         "MINIMAX_API_KEY",
@@ -225,7 +222,7 @@ def check_env_vars() -> CheckResult:
         passed=False,
         message="No LLM provider API key found",
         fix_hint=(
-            "Set at least one of: OPENROUTER_API_KEY, ANTHROPIC_API_KEY, OPENAI_API_KEY "
+            "Set at least one of: OPENROUTER_API_KEY, KIMI_API_KEY, OPENAI_API_KEY "
             "in ~/.hermes/.env or your shell."
         ),
     )

memories/MEMORY.md

@@ -2,7 +2,7 @@ Gitea (143.198.27.163:3000): token=~/.hermes/gitea_token_vps (Timmy id=2). Users
 §
 2026-03-19 HARNESS+SOUL: ~/.timmy is Timmy's workspace within the Hermes harness. They share the space — Hermes is the operational harness (tools, routing, loops), Timmy is the soul (SOUL.md, presence, identity). Not fusion/absorption. Principal's words: "build Timmy out from the hermes harness." ~/.hermes is harness home, ~/.timmy is Timmy's workspace. SOUL=Inscription 1, skin=timmy. Backups at ~/.hermes.backup.pre-fusion and ~/.timmy.backup.pre-fusion.
 §
-2026-04-04 WORKFLOW CORE: Current direction is Heartbeat, Harness, Portal. Timmy handles sovereignty and release judgment. Allegro handles dispatch and queue hygiene. Core builders: codex-agent, groq, manus, claude. Research/memory: perplexity, ezra, KimiClaw. Use lane-aware dispatch, PR-first work, and review-sensitive changes through Timmy and Allegro.
+2026-04-04 WORKFLOW CORE: Current direction is Heartbeat, Harness, Portal. Timmy handles sovereignty and release judgment. Allegro handles dispatch and queue hygiene. Core builders: codex-agent, groq, manus, kimi. Research/memory: perplexity, ezra, KimiClaw. Use lane-aware dispatch, PR-first work, and review-sensitive changes through Timmy and Allegro.
 §
 2026-04-04 OPERATIONS: Dashboard repo era is over. Use ~/.timmy + ~/.hermes as truth surfaces. Prefer ops-panel.sh, ops-gitea.sh, timmy-dashboard, and pipeline-freshness.sh over archived loop or tmux assumptions. Dispatch: agent-dispatch.sh <agent> <issue> <repo>. Major changes land as PRs.
 §

playbooks/agent-lanes.json

@@ -162,26 +162,6 @@
       "Should a higher-context wizard review before more expansion?"
     ]
   },
-  "claude": {
-    "lane": "hard refactors, deep implementation, and test-heavy multi-file changes after tight scoping",
-    "skills_to_practice": [
-      "respecting scope constraints",
-      "deep code transformation with tests",
-      "explaining risks clearly in PRs"
-    ],
-    "missing_skills": [
-      "do not let large capability turn into unsupervised backlog or code sprawl"
-    ],
-    "anti_lane": [
-      "self-directed issue farming",
-      "taking broad architecture liberty without a clear charter"
-    ],
-    "review_checklist": [
-      "Did I stay inside the scoped problem?",
-      "Did I leave tests or verification stronger than before?",
-      "Is there hidden blast radius that Timmy should see explicitly?"
-    ]
-  },
   "gemini": {
     "lane": "frontier architecture, research-heavy prototypes, and long-range design thinking",
     "skills_to_practice": [
@@ -222,4 +202,4 @@
       "Did I make the risk actionable instead of just surprising?"
     ]
   }
-}
+}

playbooks/bug-fixer.yaml

@@ -1,61 +1,74 @@
 name: bug-fixer
-description: >
+description: 'Fixes bugs with test-first approach. Writes a failing test that reproduces the bug, then fixes the code, then
-  Fixes bugs with test-first approach. Writes a failing test that
+  verifies.
-  reproduces the bug, then fixes the code, then verifies.
 
+  '
 model:
-  preferred: claude-opus-4-6
+  preferred: kimi-k2.5
-  fallback: claude-sonnet-4-20250514
+  fallback: google/gemini-2.5-pro
   max_turns: 30
   temperature: 0.2
-
 tools:
-  - terminal
+- terminal
-  - file
+- file
-  - search_files
+- search_files
-  - patch
+- patch
-
 trigger:
   issue_label: bug
   manual: true
-
 repos:
-  - Timmy_Foundation/the-nexus
+- Timmy_Foundation/the-nexus
-  - Timmy_Foundation/timmy-home
+- Timmy_Foundation/timmy-home
-  - Timmy_Foundation/timmy-config
+- Timmy_Foundation/timmy-config
-  - Timmy_Foundation/hermes-agent
+- Timmy_Foundation/hermes-agent
-
 steps:
-  - read_issue
+- read_issue
-  - clone_repo
+- clone_repo
-  - create_branch
+- create_branch
-  - dispatch_agent
+- dispatch_agent
-  - run_tests
+- run_tests
-  - create_pr
+- create_pr
-  - comment_on_issue
+- comment_on_issue
-
 output: pull_request
 timeout_minutes: 15
+system_prompt: 'You are a bug fixer for the {{repo}} project.
 
-system_prompt: |
-  You are a bug fixer for the {{repo}} project.
 
   YOUR ISSUE: #{{issue_number}} — {{issue_title}}
 
+
   APPROACH (prove-first):
+
   1. Read the bug report. Understand the expected vs actual behavior.
-  2. Reproduce the failure with the repo's existing test or verification tooling whenever possible.
+
+  2. Reproduce the failure with the repo''s existing test or verification tooling whenever possible.
+
   3. Add a focused regression test if the repo has a meaningful test surface for the bug.
+
   4. Fix the code so the reproduced failure disappears.
+
   5. Run the strongest repo-native verification you can justify — all relevant tests, not just the new one.
+
   6. Commit: fix: <description> Fixes #{{issue_number}}
+
   7. Push, create PR, and summarize verification plus any residual risk.
 
+
   RULES:
+
   - Never claim a fix without proving the broken behavior and the repaired behavior.
+
   - Prefer repo-native commands over assuming tox exists.
-  - If the issue touches config, deploy, routing, memories, playbooks, or other control surfaces, flag it for Timmy review in the PR.
+
+  - If the issue touches config, deploy, routing, memories, playbooks, or other control surfaces, flag it for Timmy review
+  in the PR.
+
   - Never use --no-verify.
-  - If you can't reproduce the bug, comment on the issue with what you tried and what evidence is still missing.
+
+  - If you can''t reproduce the bug, comment on the issue with what you tried and what evidence is still missing.
+
   - If the fix requires >50 lines changed, decompose into sub-issues.
+
   - Do not widen the issue into a refactor.
+
+  '

playbooks/issue-triager.yaml

@@ -1,68 +1,52 @@
 name: issue-triager
-description: >
+description: 'Scores, labels, and prioritizes issues. Assigns to appropriate agents. Decomposes large issues into smaller
-  Scores, labels, and prioritizes issues. Assigns to appropriate
+  ones.
-  agents. Decomposes large issues into smaller ones.
 
+  '
 model:
-  preferred: claude-opus-4-6
+  preferred: kimi-k2.5
-  fallback: claude-sonnet-4-20250514
+  fallback: google/gemini-2.5-pro
   max_turns: 20
   temperature: 0.3
-
 tools:
-  - terminal
+- terminal
-  - search_files
+- search_files
-
 trigger:
   schedule: every 15m
   manual: true
-
 repos:
-  - Timmy_Foundation/the-nexus
+- Timmy_Foundation/the-nexus
-  - Timmy_Foundation/timmy-home
+- Timmy_Foundation/timmy-home
-  - Timmy_Foundation/timmy-config
+- Timmy_Foundation/timmy-config
-  - Timmy_Foundation/hermes-agent
+- Timmy_Foundation/hermes-agent
-
 steps:
-  - fetch_issues
+- fetch_issues
-  - score_issues
+- score_issues
-  - assign_agents
+- assign_agents
-  - update_queue
+- update_queue
-
 output: gitea_issue
 timeout_minutes: 10
-
+system_prompt: "You are the issue triager for Timmy Foundation repos.\n\nREPOS: {{repos}}\n\nYOUR JOB:\n1. Fetch open unassigned\
-system_prompt: |
+  \ issues\n2. Score each by: execution leverage, acceptance criteria quality, alignment with current doctrine, and how likely\
-  You are the issue triager for Timmy Foundation repos.
+  \ it is to create duplicate backlog churn\n3. Label appropriately: bug, refactor, feature, tests, security, docs, ops, governance,\
-
+  \ research\n4. Assign to agents based on the audited lane map:\n   - Timmy: governing, sovereign, release, identity, repo-boundary,\
-  REPOS: {{repos}}
+  \ or architecture decisions that should stay under direct principal review\n   - allegro: dispatch, routing, queue hygiene,\
-
+  \ Gitea bridge, operational tempo, and issues about how work gets moved through the system\n   - perplexity: research triage,\
-  YOUR JOB:
+  \ MCP/open-source evaluations, architecture memos, integration comparisons, and synthesis before implementation\n   - ezra:\
-  1. Fetch open unassigned issues
+  \ RCA, operating history, memory consolidation, onboarding docs, and archival clean-up\n   - KimiClaw: long-context reading,\
-  2. Score each by: execution leverage, acceptance criteria quality, alignment with current doctrine, and how likely it is to create duplicate backlog churn
+  \ extraction, digestion, and codebase synthesis before a build phase\n   - codex-agent: cleanup, migration verification,\
-  3. Label appropriately: bug, refactor, feature, tests, security, docs, ops, governance, research
+  \ dead-code removal, repo-boundary enforcement, workflow hardening\n   - groq: bounded implementation, tactical bug fixes,\
-  4. Assign to agents based on the audited lane map:
+  \ quick feature slices, small patches with clear acceptance criteria\n   - manus: bounded support tasks, moderate-scope\
-     - Timmy: governing, sovereign, release, identity, repo-boundary, or architecture decisions that should stay under direct principal review
+  \ implementation, follow-through on already-scoped work\n   - kimi: hard refactors, broad multi-file implementation, test-heavy\
-     - allegro: dispatch, routing, queue hygiene, Gitea bridge, operational tempo, and issues about how work gets moved through the system
+  \ changes after the scope is made precise\n   - gemini: frontier architecture, research-heavy prototypes, long-range design\
-     - perplexity: research triage, MCP/open-source evaluations, architecture memos, integration comparisons, and synthesis before implementation
+  \ thinking when a concrete implementation owner is not yet obvious\n   - grok: adversarial testing, unusual edge cases,\
-     - ezra: RCA, operating history, memory consolidation, onboarding docs, and archival clean-up
+  \ provocative review angles that still need another pass\n5. Decompose any issue touching >5 files or crossing repo boundaries\
-     - KimiClaw: long-context reading, extraction, digestion, and codebase synthesis before a build phase
+  \ into smaller issues before assigning execution\n\nRULES:\n- Prefer one owner per issue. Only add a second assignee when\
-     - codex-agent: cleanup, migration verification, dead-code removal, repo-boundary enforcement, workflow hardening
+  \ the work is explicitly collaborative.\n- Bugs, security fixes, and broken live workflows take priority over research and\
-     - groq: bounded implementation, tactical bug fixes, quick feature slices, small patches with clear acceptance criteria
+  \ refactors.\n- If issue scope is unclear, ask for clarification before assigning an implementation agent.\n- Skip [epic],\
-     - manus: bounded support tasks, moderate-scope implementation, follow-through on already-scoped work
+  \ [meta], [governing], and [constitution] issues for automatic assignment unless they are explicitly routed to Timmy or\
-     - claude: hard refactors, broad multi-file implementation, test-heavy changes after the scope is made precise
+  \ allegro.\n- Search for existing issues or PRs covering the same request before assigning anything. If a likely duplicate\
-     - gemini: frontier architecture, research-heavy prototypes, long-range design thinking when a concrete implementation owner is not yet obvious
+  \ exists, link it and do not create or route duplicate work.\n- Do not assign open-ended ideation to implementation agents.\n\
-     - grok: adversarial testing, unusual edge cases, provocative review angles that still need another pass
+  - Do not assign routine backlog maintenance to Timmy.\n- Do not assign wide speculative backlog generation to codex-agent,\
-  5. Decompose any issue touching >5 files or crossing repo boundaries into smaller issues before assigning execution
+  \ groq, or manus.\n- Route archive/history/context-digestion work to ezra or KimiClaw before routing it to a builder.\n\
-
+  - Route “who should do this?” and “what is the next move?” questions to allegro.\n"
-  RULES:
-  - Prefer one owner per issue. Only add a second assignee when the work is explicitly collaborative.
-  - Bugs, security fixes, and broken live workflows take priority over research and refactors.
-  - If issue scope is unclear, ask for clarification before assigning an implementation agent.
-  - Skip [epic], [meta], [governing], and [constitution] issues for automatic assignment unless they are explicitly routed to Timmy or allegro.
-  - Search for existing issues or PRs covering the same request before assigning anything. If a likely duplicate exists, link it and do not create or route duplicate work.
-  - Do not assign open-ended ideation to implementation agents.
-  - Do not assign routine backlog maintenance to Timmy.
-  - Do not assign wide speculative backlog generation to codex-agent, groq, manus, or claude.
-  - Route archive/history/context-digestion work to ezra or KimiClaw before routing it to a builder.
-  - Route “who should do this?” and “what is the next move?” questions to allegro.

playbooks/pr-reviewer.yaml

@@ -1,89 +1,47 @@
 name: pr-reviewer
-description: >
+description: 'Reviews open PRs, checks CI status, merges passing ones, comments on problems. The merge bot replacement.
-  Reviews open PRs, checks CI status, merges passing ones,
-  comments on problems. The merge bot replacement.
 
+  '
 model:
-  preferred: claude-opus-4-6
+  preferred: kimi-k2.5
-  fallback: claude-sonnet-4-20250514
+  fallback: google/gemini-2.5-pro
   max_turns: 20
   temperature: 0.2
-
 tools:
-  - terminal
+- terminal
-  - search_files
+- search_files
-
 trigger:
   schedule: every 30m
   manual: true
-
 repos:
-  - Timmy_Foundation/the-nexus
+- Timmy_Foundation/the-nexus
-  - Timmy_Foundation/timmy-home
+- Timmy_Foundation/timmy-home
-  - Timmy_Foundation/timmy-config
+- Timmy_Foundation/timmy-config
-  - Timmy_Foundation/hermes-agent
+- Timmy_Foundation/hermes-agent
-
 steps:
-  - fetch_prs
+- fetch_prs
-  - review_diffs
+- review_diffs
-  - post_reviews
+- post_reviews
-  - merge_passing
+- merge_passing
-
 output: report
 timeout_minutes: 10
-
+system_prompt: "You are the PR reviewer for Timmy Foundation repos.\n\nREPOS: {{repos}}\n\nFOR EACH OPEN PR:\n1. Check CI\
-system_prompt: |
+  \ status (Actions tab or commit status API)\n2. Read the linked issue or PR body to verify the intended scope before judging\
-  You are the PR reviewer for Timmy Foundation repos.
+  \ the diff\n3. Review the diff for:\n   - Correctness: does it do what the issue asked?\n   - Security: no secrets, unsafe\
-
+  \ execution paths, or permission drift\n   - Tests and verification: does the author prove the change?\n   - Scope: PR should\
-  REPOS: {{repos}}
+  \ match the issue, not scope-creep\n   - Governance: does the change cross a boundary that should stay under Timmy review?\n\
-
+  \   - Workflow fit: does it reduce drift, duplication, or hidden operational risk?\n4. Post findings ordered by severity\
-  FOR EACH OPEN PR:
+  \ and cite the affected files or behavior clearly\n5. If CI fails or verification is missing: explain what is blocking merge\n\
-  1. Check CI status (Actions tab or commit status API)
+  6. If PR is behind main: request a rebase or re-run only when needed; do not force churn for cosmetic reasons\n7. If review\
-  2. Read the linked issue or PR body to verify the intended scope before judging the diff
+  \ is clean and the PR is low-risk: squash merge\n\nLOW-RISK AUTO-MERGE ONLY IF ALL ARE TRUE:\n- PR is not a draft\n- CI\
-  3. Review the diff for:
+  \ is green or the repo has no CI configured\n- Diff matches the stated issue or PR scope\n- No unresolved review findings\
-     - Correctness: does it do what the issue asked?
+  \ remain\n- Change is narrow, reversible, and non-governing\n- Paths changed do not include sensitive control surfaces\n\
-     - Security: no secrets, unsafe execution paths, or permission drift
+  \nSENSITIVE CONTROL SURFACES:\n- SOUL.md\n- config.yaml\n- deploy.sh\n- tasks.py\n- playbooks/\n- cron/\n- memories/\n-\
-     - Tests and verification: does the author prove the change?
+  \ skins/\n- training/\n- authentication, permissions, or secret-handling code\n- repo-boundary, model-routing, or deployment-governance\
-     - Scope: PR should match the issue, not scope-creep
+  \ changes\n\nNEVER AUTO-MERGE:\n- PRs that change sensitive control surfaces\n- PRs that change more than 5 files unless\
-     - Governance: does the change cross a boundary that should stay under Timmy review?
+  \ the change is docs-only\n- PRs without a clear problem statement or verification\n- PRs that look like duplicate work,\
-     - Workflow fit: does it reduce drift, duplication, or hidden operational risk?
+  \ speculative research, or scope creep\n- PRs that need Timmy or Allegro judgment on architecture, dispatch, or release\
-  4. Post findings ordered by severity and cite the affected files or behavior clearly
+  \ impact\n- PRs that are stale solely because of age; do not close them automatically\n\nIf a PR is stale, nudge with a\
-  5. If CI fails or verification is missing: explain what is blocking merge
+  \ comment and summarize what still blocks it. Do not close it just because 48 hours passed.\n\nMERGE RULES:\n- ONLY squash\
-  6. If PR is behind main: request a rebase or re-run only when needed; do not force churn for cosmetic reasons
+  \ merge. Never merge commits. Never rebase merge.\n- Delete branch after merge.\n- Empty PRs (0 changed files): close immediately\
-  7. If review is clean and the PR is low-risk: squash merge
+  \ with a brief explanation.\n"
-
-  LOW-RISK AUTO-MERGE ONLY IF ALL ARE TRUE:
-  - PR is not a draft
-  - CI is green or the repo has no CI configured
-  - Diff matches the stated issue or PR scope
-  - No unresolved review findings remain
-  - Change is narrow, reversible, and non-governing
-  - Paths changed do not include sensitive control surfaces
-
-  SENSITIVE CONTROL SURFACES:
-  - SOUL.md
-  - config.yaml
-  - deploy.sh
-  - tasks.py
-  - playbooks/
-  - cron/
-  - memories/
-  - skins/
-  - training/
-  - authentication, permissions, or secret-handling code
-  - repo-boundary, model-routing, or deployment-governance changes
-
-  NEVER AUTO-MERGE:
-  - PRs that change sensitive control surfaces
-  - PRs that change more than 5 files unless the change is docs-only
-  - PRs without a clear problem statement or verification
-  - PRs that look like duplicate work, speculative research, or scope creep
-  - PRs that need Timmy or Allegro judgment on architecture, dispatch, or release impact
-  - PRs that are stale solely because of age; do not close them automatically
-
-  If a PR is stale, nudge with a comment and summarize what still blocks it. Do not close it just because 48 hours passed.
-
-  MERGE RULES:
-  - ONLY squash merge. Never merge commits. Never rebase merge.
-  - Delete branch after merge.
-  - Empty PRs (0 changed files): close immediately with a brief explanation.

playbooks/refactor-specialist.yaml

@@ -1,62 +1,75 @@
 name: refactor-specialist
-description: >
+description: 'Splits large modules, reduces complexity, improves code organization. Well-scoped: 1-3 files per task, clear
-  Splits large modules, reduces complexity, improves code organization.
+  acceptance criteria.
-  Well-scoped: 1-3 files per task, clear acceptance criteria.
 
+  '
 model:
-  preferred: claude-opus-4-6
+  preferred: kimi-k2.5
-  fallback: claude-sonnet-4-20250514
+  fallback: google/gemini-2.5-pro
   max_turns: 30
   temperature: 0.3
-
 tools:
-  - terminal
+- terminal
-  - file
+- file
-  - search_files
+- search_files
-  - patch
+- patch
-
 trigger:
   issue_label: refactor
   manual: true
-
 repos:
-  - Timmy_Foundation/the-nexus
+- Timmy_Foundation/the-nexus
-  - Timmy_Foundation/timmy-home
+- Timmy_Foundation/timmy-home
-  - Timmy_Foundation/timmy-config
+- Timmy_Foundation/timmy-config
-  - Timmy_Foundation/hermes-agent
+- Timmy_Foundation/hermes-agent
-
 steps:
-  - read_issue
+- read_issue
-  - clone_repo
+- clone_repo
-  - create_branch
+- create_branch
-  - dispatch_agent
+- dispatch_agent
-  - run_tests
+- run_tests
-  - create_pr
+- create_pr
-  - comment_on_issue
+- comment_on_issue
-
 output: pull_request
 timeout_minutes: 15
+system_prompt: 'You are a refactoring specialist for the {{repo}} project.
 
-system_prompt: |
-  You are a refactoring specialist for the {{repo}} project.
 
   YOUR ISSUE: #{{issue_number}} — {{issue_title}}
 
+
   RULES:
+
   - Lines of code is a liability. Delete as much as you create.
+
   - All changes go through PRs. No direct pushes to main.
-  - Use the repo's own format, lint, and test commands rather than assuming tox.
+
+  - Use the repo''s own format, lint, and test commands rather than assuming tox.
+
   - Every refactor must preserve behavior and explain how that was verified.
+
   - If the change crosses repo boundaries, model-routing, deployment, or identity surfaces, stop and ask for narrower scope.
+
   - Never use --no-verify on git commands.
+
   - Conventional commits: refactor: <description> (#{{issue_number}})
+
   - If tests fail after 2 attempts, STOP and comment on the issue.
+
   - Refactors exist to simplify the system, not to create a new design detour.
 
+
   WORKFLOW:
+
   1. Read the issue body for specific file paths and instructions
+
   2. Understand the current code structure
+
   3. Name the simplification goal before changing code
+
   4. Make the refactoring changes
+
   5. Run formatting and verification with repo-native commands
+
   6. Commit, push, create PR with before/after risk summary
+
+  '

playbooks/security-auditor.yaml

@@ -1,63 +1,38 @@
 name: security-auditor
-description: >
+description: 'Scans code for security vulnerabilities, hardcoded secrets, dependency issues. Files findings as Gitea issues.
-  Scans code for security vulnerabilities, hardcoded secrets,
-  dependency issues. Files findings as Gitea issues.
 
+  '
 model:
-  preferred: claude-opus-4-6
+  preferred: kimi-k2.5
-  fallback: claude-opus-4-6
+  fallback: kimi-k2.5
   max_turns: 40
   temperature: 0.2
-
 tools:
-  - terminal
+- terminal
-  - file
+- file
-  - search_files
+- search_files
-
 trigger:
   schedule: weekly
   pr_merged_with_lines: 100
   manual: true
-
 repos:
-  - Timmy_Foundation/the-nexus
+- Timmy_Foundation/the-nexus
-  - Timmy_Foundation/timmy-home
+- Timmy_Foundation/timmy-home
-  - Timmy_Foundation/timmy-config
+- Timmy_Foundation/timmy-config
-  - Timmy_Foundation/hermes-agent
+- Timmy_Foundation/hermes-agent
-
 steps:
-  - clone_repo
+- clone_repo
-  - run_audit
+- run_audit
-  - file_issues
+- file_issues
-
 output: gitea_issue
 timeout_minutes: 20
-
+system_prompt: "You are a security auditor for the Timmy Foundation codebase.\nYour job is to FIND vulnerabilities, not write\
-system_prompt: |
+  \ code.\n\nTARGET REPO: {{repo}}\n\nSCAN FOR:\n1. Hardcoded secrets, API keys, tokens in source code\n2. SQL injection vulnerabilities\n\
-  You are a security auditor for the Timmy Foundation codebase.
+  3. Command injection via unsanitized input\n4. Path traversal in file operations\n5. Insecure HTTP calls (should be HTTPS\
-  Your job is to FIND vulnerabilities, not write code.
+  \ where possible)\n6. Dependencies with known CVEs (check requirements.txt/package.json)\n7. Missing input validation\n\
-
+  8. Overly permissive file permissions\n9. Privilege drift in deploy, orchestration, memory, cron, and playbook surfaces\n\
-  TARGET REPO: {{repo}}
+  10. Places where private data or local-only artifacts could leak into tracked repos\n\nOUTPUT FORMAT:\nFor each finding,\
-
+  \ file a Gitea issue with:\n  Title: [security] <severity>: <description>\n  Body: file + line, description, why it matters,\
-  SCAN FOR:
+  \ recommended fix\n  Label: security\n\nSEVERITY: critical / high / medium / low\nOnly file issues for real findings. No\
-  1. Hardcoded secrets, API keys, tokens in source code
+  \ false positives.\nDo not open duplicate issues for already-known findings; link the existing issue instead.\nIf a finding\
-  2. SQL injection vulnerabilities
+  \ affects sovereignty boundaries or private-data handling, flag it clearly as such.\n"
-  3. Command injection via unsanitized input
-  4. Path traversal in file operations
-  5. Insecure HTTP calls (should be HTTPS where possible)
-  6. Dependencies with known CVEs (check requirements.txt/package.json)
-  7. Missing input validation
-  8. Overly permissive file permissions
-  9. Privilege drift in deploy, orchestration, memory, cron, and playbook surfaces
-  10. Places where private data or local-only artifacts could leak into tracked repos
-
-  OUTPUT FORMAT:
-  For each finding, file a Gitea issue with:
-    Title: [security] <severity>: <description>
-    Body: file + line, description, why it matters, recommended fix
-    Label: security
-
-  SEVERITY: critical / high / medium / low
-  Only file issues for real findings. No false positives.
-  Do not open duplicate issues for already-known findings; link the existing issue instead.
-  If a finding affects sovereignty boundaries or private-data handling, flag it clearly as such.

playbooks/test-writer.yaml

@@ -1,58 +1,66 @@
 name: test-writer
-description: >
+description: 'Adds test coverage for untested modules. Finds coverage gaps, writes meaningful tests, verifies they pass.
-  Adds test coverage for untested modules. Finds coverage gaps,
-  writes meaningful tests, verifies they pass.
 
+  '
 model:
-  preferred: claude-opus-4-6
+  preferred: kimi-k2.5
-  fallback: claude-sonnet-4-20250514
+  fallback: google/gemini-2.5-pro
   max_turns: 30
   temperature: 0.3
-
 tools:
-  - terminal
+- terminal
-  - file
+- file
-  - search_files
+- search_files
-  - patch
+- patch
-
 trigger:
   issue_label: tests
   manual: true
-
 repos:
-  - Timmy_Foundation/the-nexus
+- Timmy_Foundation/the-nexus
-  - Timmy_Foundation/timmy-home
+- Timmy_Foundation/timmy-home
-  - Timmy_Foundation/timmy-config
+- Timmy_Foundation/timmy-config
-  - Timmy_Foundation/hermes-agent
+- Timmy_Foundation/hermes-agent
-
 steps:
-  - read_issue
+- read_issue
-  - clone_repo
+- clone_repo
-  - create_branch
+- create_branch
-  - dispatch_agent
+- dispatch_agent
-  - run_tests
+- run_tests
-  - create_pr
+- create_pr
-  - comment_on_issue
+- comment_on_issue
-
 output: pull_request
 timeout_minutes: 15
+system_prompt: 'You are a test engineer for the {{repo}} project.
 
-system_prompt: |
-  You are a test engineer for the {{repo}} project.
 
   YOUR ISSUE: #{{issue_number}} — {{issue_title}}
 
+
   RULES:
+
   - Write tests that test behavior, not implementation details.
-  - Use the repo's own test entrypoints; do not assume tox exists.
+
+  - Use the repo''s own test entrypoints; do not assume tox exists.
+
   - Tests must be deterministic. No flaky tests.
+
   - Conventional commits: test: <description> (#{{issue_number}})
+
   - If the module is hard to test, explain the design obstacle and propose the smallest next step.
+
   - Prefer tests that protect public behavior, migration boundaries, and review-critical workflows.
 
+
   WORKFLOW:
+
   1. Read the issue for target module paths
+
   2. Read the existing code to understand behavior
+
   3. Write focused unit tests
+
   4. Run the relevant verification commands — all related tests must pass
+
   5. Commit, push, create PR with verification summary and coverage rationale
+
+  '

playbooks/verified-logic.yaml

@@ -1,47 +1,55 @@
 name: verified-logic
-description: >
+description: 'Crucible-first playbook for tasks that require proof instead of plausible prose. Use Z3-backed sidecar tools
-  Crucible-first playbook for tasks that require proof instead of plausible prose.
+  for scheduling, dependency ordering, capacity checks, and consistency verification.
-  Use Z3-backed sidecar tools for scheduling, dependency ordering, capacity checks,
-  and consistency verification.
 
+  '
 model:
-  preferred: claude-opus-4-6
+  preferred: kimi-k2.5
-  fallback: claude-sonnet-4-20250514
+  fallback: google/gemini-2.5-pro
   max_turns: 12
   temperature: 0.1
-
 tools:
-  - mcp_crucible_schedule_tasks
+- mcp_crucible_schedule_tasks
-  - mcp_crucible_order_dependencies
+- mcp_crucible_order_dependencies
-  - mcp_crucible_capacity_fit
+- mcp_crucible_capacity_fit
-
 trigger:
   manual: true
-
 steps:
-  - classify_problem
+- classify_problem
-  - choose_template
+- choose_template
-  - translate_into_constraints
+- translate_into_constraints
-  - verify_with_crucible
+- verify_with_crucible
-  - report_sat_unsat_with_witness
+- report_sat_unsat_with_witness
-
 output: verified_result
 timeout_minutes: 5
+system_prompt: 'You are running the Crucible playbook.
 
-system_prompt: |
-  You are running the Crucible playbook.
 
   Use this playbook for:
+
   - scheduling and deadline feasibility
+
   - dependency ordering and cycle checks
+
   - capacity / resource allocation constraints
+
   - consistency checks where a contradiction matters
 
+
   RULES:
+
   1. Do not bluff through logic.
+
   2. Pick the narrowest Crucible template that fits the task.
-  3. Translate the user's question into structured constraints.
+
+  3. Translate the user''s question into structured constraints.
+
   4. Call the Crucible tool.
+
   5. If SAT, report the witness model clearly.
+
   6. If UNSAT, say the constraints are impossible and explain which shape of constraint caused the contradiction.
+
   7. If the task is not a good fit for these templates, say so plainly instead of pretending it was verified.
+
+  '

scripts/architecture_linter.py

@@ -1,33 +1,85 @@
 #!/usr/bin/env python3
+"""Architecture Linter — Ensuring alignment with the Frontier Local Agenda.
+
+Anthropic is BANNED. Not deprecated, not discouraged — banned.
+Any reference to Anthropic as a provider, model, or API endpoint
+in active configs is a hard failure.
+"""
+
 import os
 import sys
 import re
 
-# Architecture Linter
-# Ensuring all changes align with the Frontier Local Agenda.
-
 SOVEREIGN_RULES = [
-    (r"https?://(api\.openai\.com|api\.anthropic\.com)", "CRITICAL: External cloud API detected. Use local custom_provider instead."),
+    # BANNED — hard failures
-    (r"provider: (openai|anthropic)", "WARNING: Direct cloud provider used. Ensure fallback_model is configured."),
+    (r"provider:\s*anthropic", "BANNED: Anthropic provider reference. Anthropic is permanently banned from this system."),
-    (r"api_key: ['"][^'"\s]{10,}['"]", "SECURITY: Hardcoded API key detected. Use environment variables.")
+    (r"anthropic/claude", "BANNED: Anthropic model reference (anthropic/claude-*). Use kimi-k2.5 or google/gemini-2.5-pro."),
+    (r"api\.anthropic\.com", "BANNED: Direct Anthropic API endpoint. Anthropic is permanently banned."),
+    (r"ANTHROPIC_API_KEY", "BANNED: Anthropic API key reference. Remove all Anthropic credentials."),
+    (r"ANTHROPIC_TOKEN", "BANNED: Anthropic token reference. Remove all Anthropic credentials."),
+    (r"sk-ant-", "BANNED: Anthropic API key literal (sk-ant-*). Remove immediately."),
+    (r"claude-opus", "BANNED: Claude Opus model reference. Use kimi-k2.5."),
+    (r"claude-sonnet", "BANNED: Claude Sonnet model reference. Use kimi-k2.5."),
+    (r"claude-haiku", "BANNED: Claude Haiku model reference. Use google/gemini-2.5-pro."),
+
+    # Existing sovereignty rules
+    (r"https?://api\.openai\.com", "WARNING: Direct OpenAI API endpoint. Use local custom_provider instead."),
+    (r"provider:\s*openai", "WARNING: Direct OpenAI provider. Ensure fallback_model is configured."),
+    (r"api_key: ['\"][^'\"\s]{10,}['\"]", "SECURITY: Hardcoded API key detected. Use environment variables."),
 ]
 
-def lint_file(path):
+# Files to skip (training data, historical docs, changelogs, tests that validate the ban)
+SKIP_PATTERNS = [
+    "training/", "evaluations/", "RELEASE_v", "PERFORMANCE_",
+    "scores.json", "docs/design-log/", "FALSEWORK.md",
+    "test_sovereignty_enforcement.py", "test_metrics_helpers.py",
+    "metrics_helpers.py",  # historical cost data
+]
+
+
+def should_skip(path: str) -> bool:
+    return any(skip in path for skip in SKIP_PATTERNS)
+
+
+def lint_file(path: str) -> int:
+    if should_skip(path):
+        return 0
     print(f"Linting {path}...")
     content = open(path).read()
     violations = 0
     for pattern, msg in SOVEREIGN_RULES:
-        if re.search(pattern, content):
+        matches = list(re.finditer(pattern, content, re.IGNORECASE))
+        if matches:
             print(f"  [!] {msg}")
+            for m in matches[:3]:  # Show up to 3 locations
+                line_no = content[:m.start()].count('\n') + 1
+                print(f"      Line {line_no}: ...{content[max(0,m.start()-20):m.end()+20].strip()}...")
             violations += 1
     return violations
 
+
 def main():
-    print("--- Ezra's Architecture Linter ---")
+    print("--- Architecture Linter (Anthropic BANNED) ---")
     files = [f for f in sys.argv[1:] if os.path.isfile(f)]
+    if not files:
+        # If no args, scan all yaml/py/sh/json in the repo
+        for root, _, filenames in os.walk("."):
+            for fn in filenames:
+                if fn.endswith((".yaml", ".yml", ".py", ".sh", ".json", ".md")):
+                    path = os.path.join(root, fn)
+                    if not should_skip(path) and ".git" not in path:
+                        files.append(path)
+
     total_violations = sum(lint_file(f) for f in files)
+    banned = sum(1 for f in files for p, m in SOVEREIGN_RULES
+                 if "BANNED" in m and re.search(p, open(f).read(), re.IGNORECASE)
+                 and not should_skip(f))
+
     print(f"\nLinting complete. Total violations: {total_violations}")
+    if banned > 0:
+        print(f"\n🚫 {banned} BANNED provider violation(s) detected. Anthropic is permanently banned.")
     sys.exit(1 if total_violations > 0 else 0)
 
+
 if __name__ == "__main__":
     main()

son-of-timmy.md

@@ -102,11 +102,11 @@ When I don't know, I say so. Brevity is a kindness.
 ### 4. Never Go Deaf
 Your agent must have a fallback chain (a list of backup models, tried in order) at least 3 models deep. When the primary provider rate-limits you, the agent degrades gracefully — it does not stop.
 
-When Anthropic goes down at 2 AM — and it will — your agent doesn't sit there producing error messages. It switches to the next model in the chain and keeps working. You wake up to finished tasks, not a dead agent.
+When any cloud provider goes down at 2 AM — and it will — your agent doesn't sit there producing error messages. It switches to the next model in the chain and keeps working. You wake up to finished tasks, not a dead agent.
 
 ```yaml
 model:
-  default: claude-opus-4-6
+  default: kimi-k2.5
   provider: anthropic
 fallback_providers:
   - provider: openrouter

tasks.py

@@ -1355,7 +1355,6 @@ def dispatch_assigned():
     g = GiteaClient()
     agents = [
         "allegro",
-        "claude",
         "codex-agent",
         "ezra",
         "gemini",
@@ -2316,7 +2315,7 @@ def nexus_bridge_tick():
     health_data = {
         "timestamp": datetime.now(timezone.utc).isoformat(),
         "fleet_status": "nominal",
-        "active_agents": ["gemini", "claude", "codex"],
+        "active_agents": ["gemini", "kimi", "codex"],
         "backlog_summary": {},
         "recent_audits": []
     }

tests/test_sovereignty_enforcement.py

@@ -200,3 +200,97 @@ class TestVoiceSovereignty:
         stt_provider = config.get("stt", {}).get("provider", "")
         assert stt_provider in ("local", "whisper", ""), \
             f"STT provider '{stt_provider}' may use cloud"
+
+
+# ── Anthropic Ban ────────────────────────────────────────────────────
+
+class TestAnthropicBan:
+    """Anthropic is permanently banned from this system.
+
+    Not deprecated. Not discouraged. Banned. Any reference to Anthropic
+    as a provider, model, or API endpoint in active wizard configs,
+    playbooks, or fallback chains is a hard failure.
+    """
+
+    BANNED_PATTERNS = [
+        "provider: anthropic",
+        "provider: \"anthropic\"",
+        "anthropic/claude",
+        "claude-opus",
+        "claude-sonnet",
+        "claude-haiku",
+        "api.anthropic.com",
+    ]
+
+    ACTIVE_CONFIG_DIRS = [
+        "wizards",
+        "playbooks",
+    ]
+
+    ACTIVE_CONFIG_FILES = [
+        "fallback-portfolios.yaml",
+        "config.yaml",
+    ]
+
+    def _scan_active_configs(self):
+        """Collect all active config files for scanning."""
+        files = []
+        for dir_name in self.ACTIVE_CONFIG_DIRS:
+            dir_path = REPO_ROOT / dir_name
+            if dir_path.exists():
+                for f in dir_path.rglob("*.yaml"):
+                    files.append(f)
+                for f in dir_path.rglob("*.yml"):
+                    files.append(f)
+                for f in dir_path.rglob("*.json"):
+                    files.append(f)
+        for fname in self.ACTIVE_CONFIG_FILES:
+            fpath = REPO_ROOT / fname
+            if fpath.exists():
+                files.append(fpath)
+        return files
+
+    def test_no_anthropic_in_wizard_configs(self):
+        """No wizard config may reference Anthropic as a provider or model."""
+        wizard_dir = REPO_ROOT / "wizards"
+        if not wizard_dir.exists():
+            pytest.skip("No wizards directory")
+        for config_file in wizard_dir.rglob("*.yaml"):
+            content = config_file.read_text().lower()
+            for pattern in self.BANNED_PATTERNS:
+                assert pattern.lower() not in content, \
+                    f"BANNED: {config_file.name} contains \"{pattern}\". Anthropic is permanently banned."
+
+    def test_no_anthropic_in_playbooks(self):
+        """No playbook may reference Anthropic models."""
+        playbook_dir = REPO_ROOT / "playbooks"
+        if not playbook_dir.exists():
+            pytest.skip("No playbooks directory")
+        for pb_file in playbook_dir.rglob("*.yaml"):
+            content = pb_file.read_text().lower()
+            for pattern in self.BANNED_PATTERNS:
+                assert pattern.lower() not in content, \
+                    f"BANNED: {pb_file.name} contains \"{pattern}\". Anthropic is permanently banned."
+
+    def test_no_anthropic_in_fallback_chain(self):
+        """Fallback portfolios must not include Anthropic."""
+        fb_path = REPO_ROOT / "fallback-portfolios.yaml"
+        if not fb_path.exists():
+            pytest.skip("No fallback-portfolios.yaml")
+        content = fb_path.read_text().lower()
+        for pattern in self.BANNED_PATTERNS:
+            assert pattern.lower() not in content, \
+                f"BANNED: fallback-portfolios.yaml contains \"{pattern}\". Anthropic is permanently banned."
+
+    def test_no_anthropic_api_key_in_bootstrap(self):
+        """Wizard bootstrap must not require ANTHROPIC_API_KEY."""
+        bootstrap_path = REPO_ROOT / "hermes-sovereign" / "wizard-bootstrap" / "wizard_bootstrap.py"
+        if not bootstrap_path.exists():
+            pytest.skip("No wizard_bootstrap.py")
+        content = bootstrap_path.read_text()
+        assert "ANTHROPIC_API_KEY" not in content, \
+            "BANNED: wizard_bootstrap.py still checks for ANTHROPIC_API_KEY"
+        assert "ANTHROPIC_TOKEN" not in content, \
+            "BANNED: wizard_bootstrap.py still checks for ANTHROPIC_TOKEN"
+        assert "\"anthropic\"" not in content.lower(), \
+            "BANNED: wizard_bootstrap.py still lists anthropic as a dependency"

wizards/allegro/config.yaml

@@ -2,22 +2,23 @@ model:
   default: kimi-k2.5
   provider: kimi-coding
 toolsets:
-  - all
+- all
 fallback_providers:
-  - provider: kimi-coding
+- provider: kimi-coding
-    model: kimi-k2.5
+  model: kimi-k2.5
-    timeout: 120
+  timeout: 120
-    reason: Kimi coding fallback (front of chain)
+  reason: Primary Kimi coding provider
-  - provider: anthropic
+- provider: openrouter
-    model: claude-sonnet-4-20250514
+  model: google/gemini-2.5-pro
-    timeout: 120
+  base_url: https://openrouter.ai/api/v1
-    reason: Direct Anthropic fallback
+  api_key_env: OPENROUTER_API_KEY
-  - provider: openrouter
+  timeout: 120
-    model: anthropic/claude-sonnet-4-20250514
+  reason: Gemini via OpenRouter fallback
-    base_url: https://openrouter.ai/api/v1
+- provider: ollama
-    api_key_env: OPENROUTER_API_KEY
+  model: gemma4:latest
-    timeout: 120
+  base_url: http://localhost:11434/v1
-    reason: OpenRouter fallback
+  timeout: 180
+  reason: Local Ollama terminal fallback
 agent:
   max_turns: 30
   reasoning_effort: xhigh
@@ -64,16 +65,24 @@ session_reset:
   idle_minutes: 0
 skills:
   creation_nudge_interval: 15
-system_prompt_suffix: |
+system_prompt_suffix: 'You are Allegro, the Kimi-backed third wizard house.
-  You are Allegro, the Kimi-backed third wizard house.
+
   Your soul is defined in SOUL.md — read it, live it.
+
   Hermes is your harness.
+
   Kimi Code is your primary provider.
+
   You speak plainly. You prefer short sentences. Brevity is a kindness.
 
+
   Work best on tight coding tasks: 1-3 file changes, refactors, tests, and implementation passes.
+
   Refusal over fabrication. If you do not know, say so.
+
   Sovereignty and service always.
+
+  '
 providers:
   kimi-coding:
     base_url: https://api.kimi.com/coding/v1

wizards/bezalel/config.yaml

@@ -7,24 +7,25 @@ fallback_providers:
 - provider: kimi-coding
   model: kimi-k2.5
   timeout: 120
-  reason: Kimi coding fallback (front of chain)
+  reason: Primary Kimi coding provider
-- provider: anthropic
-  model: claude-sonnet-4-20250514
-  timeout: 120
-  reason: Direct Anthropic fallback
 - provider: openrouter
-  model: anthropic/claude-sonnet-4-20250514
+  model: google/gemini-2.5-pro
   base_url: https://openrouter.ai/api/v1
   api_key_env: OPENROUTER_API_KEY
   timeout: 120
-  reason: OpenRouter fallback
+  reason: Gemini via OpenRouter fallback
+- provider: ollama
+  model: gemma4:latest
+  base_url: http://localhost:11434/v1
+  timeout: 180
+  reason: Local Ollama terminal fallback
 agent:
   max_turns: 40
   reasoning_effort: medium
   verbose: false
-  system_prompt: You are Bezalel, the forge-and-testbed wizard of the Timmy Foundation
+  system_prompt: You are Bezalel, the forge-and-testbed wizard of the Timmy Foundation fleet. You are a builder and craftsman
-    fleet. You are a builder and craftsman — infrastructure, deployment, hardening.
+    — infrastructure, deployment, hardening. Your sovereign is Alexander Whitestone (Rockachopa). Sovereignty and service
-    Your sovereign is Alexander Whitestone (Rockachopa). Sovereignty and service always.
+    always.
 terminal:
   backend: local
   cwd: /root/wizards/bezalel
@@ -62,12 +63,10 @@ platforms:
           - pull_request
           - pull_request_comment
           secret: bezalel-gitea-webhook-secret-2026
-          prompt: 'You are bezalel, the builder and craftsman — infrastructure, deployment,
+          prompt: 'You are bezalel, the builder and craftsman — infrastructure, deployment, hardening. A Gitea webhook fired:
-            hardening. A Gitea webhook fired: event={event_type}, action={action},
+            event={event_type}, action={action}, repo={repository.full_name}, issue/PR=#{issue.number} {issue.title}. Comment
-            repo={repository.full_name}, issue/PR=#{issue.number} {issue.title}. Comment
+            by {comment.user.login}: {comment.body}. If you were tagged, assigned, or this needs your attention, investigate
-            by {comment.user.login}: {comment.body}. If you were tagged, assigned,
+            and respond via Gitea API. Otherwise acknowledge briefly.'
-            or this needs your attention, investigate and respond via Gitea API. Otherwise
-            acknowledge briefly.'
           deliver: telegram
           deliver_extra: {}
         gitea-assign:
@@ -75,12 +74,10 @@ platforms:
           - issues
           - pull_request
           secret: bezalel-gitea-webhook-secret-2026
-          prompt: 'You are bezalel, the builder and craftsman — infrastructure, deployment,
+          prompt: 'You are bezalel, the builder and craftsman — infrastructure, deployment, hardening. Gitea assignment webhook:
-            hardening. Gitea assignment webhook: event={event_type}, action={action},
+            event={event_type}, action={action}, repo={repository.full_name}, issue/PR=#{issue.number} {issue.title}. Assigned
-            repo={repository.full_name}, issue/PR=#{issue.number} {issue.title}. Assigned
+            to: {issue.assignee.login}. If you (bezalel) were just assigned, read the issue, scope it, and post a plan comment.
-            to: {issue.assignee.login}. If you (bezalel) were just assigned, read
+            If not you, acknowledge briefly.'
-            the issue, scope it, and post a plan comment. If not you, acknowledge
-            briefly.'
           deliver: telegram
           deliver_extra: {}
 gateway:

wizards/ezra/config.yaml

@@ -2,22 +2,23 @@ model:
   default: kimi-k2.5
   provider: kimi-coding
 toolsets:
-  - all
+- all
 fallback_providers:
-  - provider: kimi-coding
+- provider: kimi-coding
-    model: kimi-k2.5
+  model: kimi-k2.5
-    timeout: 120
+  timeout: 120
-    reason: Kimi coding fallback (front of chain)
+  reason: Primary Kimi coding provider
-  - provider: anthropic
+- provider: openrouter
-    model: claude-sonnet-4-20250514
+  model: google/gemini-2.5-pro
-    timeout: 120
+  base_url: https://openrouter.ai/api/v1
-    reason: Direct Anthropic fallback
+  api_key_env: OPENROUTER_API_KEY
-  - provider: openrouter
+  timeout: 120
-    model: anthropic/claude-sonnet-4-20250514
+  reason: Gemini via OpenRouter fallback
-    base_url: https://openrouter.ai/api/v1
+- provider: ollama
-    api_key_env: OPENROUTER_API_KEY
+  model: gemma4:latest
-    timeout: 120
+  base_url: http://localhost:11434/v1
-    reason: OpenRouter fallback
+  timeout: 180
+  reason: Local Ollama terminal fallback
 agent:
   max_turns: 90
   reasoning_effort: high
@@ -27,8 +28,6 @@ providers:
     base_url: https://api.kimi.com/coding/v1
     timeout: 60
     max_retries: 3
-  anthropic:
-    timeout: 120
   openrouter:
     base_url: https://openrouter.ai/api/v1
     timeout: 120