fix: comprehensive bugfixes — 53 issues resolved

skills/ai-peer/SKILL.md

@@ -10,7 +10,7 @@ Peer-to-peer conversation rooms for AI CLI tools (Claude Code, Codex, OpenCode)
 - **Auth**: Local daemon: none. Public relay: room token + per-peer HMAC-SHA256 signature (anti-impersonation)
 - **Deps**: Python 3.10+ (stdlib only, zero pip deps)
 - **Relay**: `https://relay.ai-peer.chat` (Durable Objects, strong consistency)
-- **Operations**: 16 (daemon 3, room 4, chat 2, invite 1, quick 1, discuss 1, peer 3, export 1)
+- **Operations**: 16 sub-commands across 12 top-level commands (daemon: start/stop/status, room: create/join/list/delete, chat: send/read, invite, quick, discuss, discover, list, register, history, export, identity)
 - **Storage**: Local SQLite at `~/.ai-peers/peers.db`, daemon config at `~/.ai-peers/daemon.json`
 - **Tests**: 69 (unit + integration + crypto + auth)
 

skills/ai-peer/references/PRD.md

@@ -72,7 +72,7 @@ The closest attempt (claude-peers-mcp) is single-machine, Claude-only, and requi
 | Component | Tech | Purpose |
 |-----------|------|---------|
 | **Daemon** | Python stdlib `http.server` + SQLite | Local message store + HTTP API for CLI commands |
-| **CLI** | Python `argparse` | 16 commands: daemon, room, chat, invite, quick, discuss, discover, etc. |
+| **CLI** | Python manual dispatch | 12 top-level commands (daemon, room, chat, invite, quick, discuss, discover, list, register, history, export, identity) with sub-commands |
 | **Relay** | Cloudflare Worker + Durable Objects | Cross-internet message forwarding, strong consistency |
 | **Crypto** | PBKDF2 + Fernet (`cryptography` package) | Optional E2E encryption, relay stores only ciphertext |
 | **Spawner** | `subprocess.run` | Invoke AI CLI tools with conversation context as prompt |
@@ -186,7 +186,7 @@ The core differentiator — bring any AI CLI tool into a conversation.
 ### P1: Quality & Reliability
 - SQLite WAL mode + threading locks for concurrent access
 - Cross-platform signal handling (SIGTERM/SIGBREAK)
-- 40 automated tests (unit + integration + crypto)
+- 69 automated tests (unit + integration + crypto + auth)
 - Grade A on skill-optimizer (40/40)
 - 3-AI deep review (Claude+Codex+OpenCode), 10 HIGH issues fixed
 

skills/ai-peer/scripts/ai_peer/__init__.py

@@ -1,17 +1,29 @@
 """AI Peer — decentralized AI-to-AI and human-AI communication."""
 import json
 import sys
+import os
+
+# Force UTF-8 output on Windows to handle Unicode characters in AI responses
+if sys.platform == "win32":
+    os.environ.setdefault("PYTHONIOENCODING", "utf-8")
+    for stream_name in ("stdout", "stderr"):
+        stream = getattr(sys, stream_name, None)
+        if stream and hasattr(stream, "reconfigure"):
+            try:
+                stream.reconfigure(encoding="utf-8")
+            except Exception:
+                pass
 
 
 def output(ok, data=None, error=None):
-    """Standard output envelope."""
+    """Standard output envelope. Returns exit code."""
     result = {"ok": ok, "schema_version": "1"}
     if ok:
         result["data"] = data
     else:
         result["error"] = str(error)
     print(json.dumps(result, ensure_ascii=False, indent=2))
-    sys.exit(0 if ok else 1)
+    return 0 if ok else 1
 
 
 def main():
@@ -38,11 +50,11 @@ def main():
     try:
         from .ops import run_command
         result = run_command(args)
-        output(True, result)
+        sys.exit(output(True, result))
     except KeyboardInterrupt:
         sys.exit(130)
     except Exception as e:
-        output(False, error=e)
+        sys.exit(output(False, error=e))
 
 
 if __name__ == "__main__":

skills/ai-peer/scripts/ai_peer/client.py

@@ -29,11 +29,11 @@ def _req(self, method, path, data=None, timeout=30):
             except Exception:
                 return {"error": f"HTTP {e.code}"}
         except (urllib.error.URLError, OSError):
-            return None  # Daemon not running or connection failed
+            return {"error": "daemon not running or connection failed"}
 
     def is_alive(self):
         r = self._req("GET", "/health", timeout=3)
-        return r is not None and r.get("ok")
+        return isinstance(r, dict) and r.get("ok")
 
     # === Rooms ===
 
@@ -142,7 +142,12 @@ def ensure_daemon(host=None, port=None):
             # Need to restart with new config
             sys.stderr.write(f"⟳ Restarting daemon ({cfg_host}:{cfg_port} → {host}:{port})...\n")
             _stop_daemon()
-            time.sleep(0.5)
+            # Wait for old daemon to fully stop before starting new one
+            for _ in range(20):
+                time.sleep(0.2)
+                check_client = PeerClient(check_host, port)
+                if not check_client.is_alive():
+                    break
         else:
             return client
 
@@ -187,6 +192,13 @@ def _start_daemon(host, port):
 
     PEERS_HOME.mkdir(parents=True, exist_ok=True)
 
+    # Rotate log if too large (>1MB)
+    if LOG_FILE.exists() and LOG_FILE.stat().st_size > 1_000_000:
+        rotated = LOG_FILE.with_suffix(".log.1")
+        if rotated.exists():
+            rotated.unlink()
+        LOG_FILE.rename(rotated)
+
     scripts_dir = str(Path(__file__).resolve().parent.parent)
     existing_pypath = os.environ.get("PYTHONPATH", "")
     pypath = scripts_dir + os.pathsep + existing_pypath if existing_pypath else scripts_dir

skills/ai-peer/scripts/ai_peer/constants.py

@@ -33,12 +33,19 @@
         "env_unset": ["CLAUDECODE"],
     },
     "opencode": {
-        "cmd": ["opencode", "-p", "{prompt}"],
+        "cmd": ["opencode", "run", "{prompt}"],
         "env_unset": [],
     },
 }
 
 
 def get_machine_id():
-    """Stable machine identifier."""
-    return platform.node()
+    """Stable machine identifier with fallback chain."""
+    name = platform.node()
+    if not name:
+        try:
+            import socket
+            name = socket.gethostname()
+        except Exception:
+            pass
+    return name or "unknown"

skills/ai-peer/scripts/ai_peer/crypto.py

@@ -50,10 +50,21 @@ def decrypt(ciphertext: str, password: str, room_id: str, strict: bool = False)
     f = Fernet(key)
     try:
         return f.decrypt(ciphertext.encode()).decode()
-    except Exception:
+    except InvalidToken:
         return "[encrypted — wrong password or corrupted]"
 
 
 def is_encrypted(text: str) -> bool:
-    """Check if text looks like a Fernet token."""
-    return text.startswith("gAAA") and len(text) > 50
+    """Check if text looks like a Fernet token by attempting to decode it."""
+    if not text or not isinstance(text, str):
+        return False
+    if not HAS_CRYPTO:
+        return text.startswith("gAAA") and len(text) > 50
+    try:
+        # Fernet tokens are URL-safe base64, decode to check structure
+        import base64
+        decoded = base64.urlsafe_b64decode(text + "==")
+        # Fernet tokens start with version byte (0x80) + 8-byte timestamp
+        return len(decoded) > 9 and decoded[0] == 0x80
+    except Exception:
+        return False

skills/ai-peer/scripts/ai_peer/db.py

@@ -81,6 +81,8 @@ def __init__(self, db_path: Path):
         self.conn.executescript(SCHEMA)
         self._run_migrations()
         self._lock = threading.Lock()
+        # WAL mode allows concurrent reads without locks — only writes need the lock.
+        # Reads (get_room, get_messages, list_rooms, room_peers) are safe without _lock.
 
     def _run_migrations(self):
         for sql in MIGRATIONS:
@@ -115,7 +117,7 @@ def create_room(self, name, mode="local", relay_url=None, password=None, room_id
                     (rid, name, mode, relay_url, password, token, _now()),
                 )
             self.conn.commit()
-        return _row_to_dict(self.conn.execute("SELECT * FROM rooms WHERE id=?", (rid,)).fetchone())
+            return _row_to_dict(self.conn.execute("SELECT * FROM rooms WHERE id=?", (rid,)).fetchone())
 
     def get_room(self, room_id):
         return _row_to_dict(self.conn.execute("SELECT * FROM rooms WHERE id=?", (room_id,)).fetchone())

skills/ai-peer/scripts/ai_peer/helpers.py

@@ -143,9 +143,9 @@ def _merge_messages(local_resp, relay_resp):
             seen_sys.add(sys_key)
         elif not msg_id:
             # Non-system without ID: use content hash as fallback
-            fallback = hashlib.md5(
+            fallback = hashlib.sha256(
                 f"{msg.get('peer_id', '')}:{msg.get('content', '')}:{msg.get('created_at', '')}".encode()
-            ).hexdigest()
+            ).hexdigest()[:16]
             if fallback in seen_ids:
                 continue
             seen_ids.add(fallback)
@@ -186,7 +186,13 @@ def _peer_signature(peer_secret, room_id):
 
 def get_or_create_identity():
     """Get or create local human identity. Auto-migrates old PID-based IDs."""
-    username = os.environ.get("USER", os.environ.get("USERNAME", "user"))
+    username = os.environ.get("USER", os.environ.get("USERNAME"))
+    if not username:
+        try:
+            import getpass
+            username = getpass.getuser()
+        except Exception:
+            username = "user"
     expected_id = f"{get_machine_id()}-human-{username}"
 
     if IDENTITY_FILE.exists():

skills/ai-peer/scripts/ai_peer/ops_ai.py

@@ -5,7 +5,7 @@
 import secrets
 import sys
 
-from .constants import get_machine_id
+from .constants import get_machine_id, TOOL_COMMANDS
 from .client import ensure_daemon
 from .relay_client import RelayClient
 from .spawn import build_conversation_prompt, spawn_ai
@@ -16,6 +16,14 @@
 )
 
 
+def _safe_int(value, name):
+    """Parse int with clear error message."""
+    try:
+        return int(value)
+    except (ValueError, TypeError):
+        raise ValueError(f"Invalid {name}: {value!r} (must be a number)")
+
+
 def cmd_invite(args):
     if not args:
         raise ValueError(
@@ -34,7 +42,7 @@ def cmd_invite(args):
         elif args[i] == "--context" and i + 1 < len(args):
             context = args[i + 1]; i += 2
         elif args[i] == "--timeout" and i + 1 < len(args):
-            timeout = int(args[i + 1]); i += 2
+            timeout = _safe_int(args[i + 1], "timeout"); i += 2
         elif not tool:
             tool = args[i]; i += 1
         else:
@@ -49,7 +57,10 @@ def cmd_invite(args):
     identity = get_or_create_identity()
     human_secret = identity.get("peer_secret", "")
     # Derive per-AI secret from human secret + tool name (true per-peer auth)
-    ai_secret = hmac.new(human_secret.encode(), tool.encode(), hashlib.sha256).hexdigest() if human_secret else ""
+    if human_secret and len(human_secret) >= 16:
+        ai_secret = hmac.new(human_secret.encode(), tool.encode(), hashlib.sha256).hexdigest()
+    else:
+        ai_secret = ""
     ai_peer_id = make_ai_peer_id(tool)
     client.register_peer(ai_peer_id, tool, "ai", tool, get_machine_id())
     client.join_room(room_id, ai_peer_id)
@@ -93,7 +104,9 @@ def cmd_invite(args):
         peer_name=tool, peer_tool=tool, peer_signature=sig,
     )
 
-    mentions = re.findall(r"@(claude-code|codex|opencode)\b", response, re.IGNORECASE)
+    # Build dynamic @mention regex from known tools
+    tool_names = "|".join(re.escape(t) for t in TOOL_COMMANDS)
+    mentions = re.findall(rf"@({tool_names})\b", response, re.IGNORECASE)
     mentions = list(dict.fromkeys(m.lower() for m in mentions if m.lower() != tool.lower()))
 
     result = {"tool": tool, "response": response, "room": room_id}
@@ -126,7 +139,7 @@ def cmd_quick(args):
         elif args[i] == "--name" and i + 1 < len(args):
             room_name = args[i + 1]; i += 2
         elif args[i] == "--timeout" and i + 1 < len(args):
-            timeout = int(args[i + 1]); i += 2
+            timeout = _safe_int(args[i + 1], "timeout"); i += 2
         elif not question and not args[i].startswith("-"):
             question = args[i]; i += 1
         else:
@@ -168,8 +181,8 @@ def cmd_quick(args):
     for tool in tools:
         sys.stderr.write(f"⏳ Inviting {tool}...\n")
         try:
-            result = cmd_invite(["--tool", tool, "--room", room_id,
-                                "--timeout", str(timeout)])
+            invite_args = ["--tool", tool, "--room", room_id, "--timeout", str(timeout)]
+            result = cmd_invite(invite_args)
             responses[tool] = result.get("response", "")
             sys.stderr.write(f"✓ {tool} responded ({len(responses[tool])} chars)\n")
         except Exception as e:
@@ -190,11 +203,11 @@ def cmd_discuss(args):
     """Run a multi-round discussion between AIs."""
     if not args:
         raise ValueError(
-            "Usage: peer discuss --tools <t1,t2> --rounds N [--room <id>] [--context \"topic\"]\n"
+            "Usage: peer discuss --tools <t1,t2> --rounds N [--room <id>] [--name <name>] [--context \"topic\"]\n"
             "  Without --room, creates a new room automatically."
         )
 
-    room_id = tools_csv = context = None
+    room_id = tools_csv = context = room_name = None
     rounds = 2
     timeout = 120
     i = 0
@@ -204,11 +217,13 @@ def cmd_discuss(args):
         elif args[i] == "--tools" and i + 1 < len(args):
             tools_csv = args[i + 1]; i += 2
         elif args[i] == "--rounds" and i + 1 < len(args):
-            rounds = int(args[i + 1]); i += 2
+            rounds = _safe_int(args[i + 1], "rounds"); i += 2
         elif args[i] == "--context" and i + 1 < len(args):
             context = args[i + 1]; i += 2
         elif args[i] == "--timeout" and i + 1 < len(args):
-            timeout = int(args[i + 1]); i += 2
+            timeout = _safe_int(args[i + 1], "timeout"); i += 2
+        elif args[i] == "--name" and i + 1 < len(args):
+            room_name = args[i + 1]; i += 2
         else:
             i += 1
 
@@ -222,7 +237,7 @@ def cmd_discuss(args):
     client = ensure_daemon()
 
     if not room_id:
-        room = client.create_room("discussion", "local")
+        room = client.create_room(room_name or "discussion", "local")
         if not room or room.get("error"):
             raise RuntimeError(f"Failed to create room: {(room or {}).get('error', 'daemon unreachable')}")
         room_id = room["id"]

skills/ai-peer/scripts/ai_peer/ops_chat.py

@@ -37,7 +37,11 @@ def cmd_chat(args):
         elif rest[i] in ("--interactive", "-i"):
             interactive = True; i += 1
         elif rest[i] == "-n" and i + 1 < len(rest):
-            limit = int(rest[i + 1]); i += 2
+            try:
+                limit = int(rest[i + 1])
+            except ValueError:
+                raise ValueError(f"Invalid number for -n: {rest[i + 1]}")
+            i += 2
         else:
             message_parts.append(rest[i]); i += 1
 
@@ -108,7 +112,7 @@ def _display_msg(msg):
 
     def _on_ws_message(data):
         if data.get("type") == "message" and data.get("data"):
-            msg = data["data"]
+            msg = dict(data["data"])  # Copy to avoid mutating shared data
             # WS delivers raw ciphertext — decrypt here (poll path decrypts in _read_messages)
             if password and msg.get("type") != "system":
                 msg["content"] = decrypt(msg["content"], password, room_id, strict=True)
@@ -136,8 +140,8 @@ def _on_ws_close():
             )
             ws_client.on_close = _on_ws_close
             ws_connected[0] = True
-        except Exception:
-            pass
+        except Exception as e:
+            sys.stderr.write(f"  WebSocket unavailable: {e}\n")
 
     # Always fetch initial context regardless of WS status
     try:
@@ -241,5 +245,6 @@ def poll():
         running[0] = False
         if ws_client:
             ws_client.close()
+        poller.join(timeout=2)
         sys.stderr.write("\nLeft the room.\n")
         return {"status": "exited"}