feat(computer-use): cua-driver backend, universal any-model schema

Background macOS desktop control via cua-driver MCP — does NOT steal the
user's cursor or keyboard focus, works with any tool-capable model.

Replaces the Anthropic-native `computer_20251124` approach from the
abandoned #4562 with a generic OpenAI function-calling schema plus SOM
(set-of-mark) captures so Claude, GPT, Gemini, and open models can all
drive the desktop via numbered element indices.

- `tools/computer_use/` package — swappable ComputerUseBackend ABC +
  CuaDriverBackend (stdio MCP client to trycua/cua's cua-driver binary).
- Universal `computer_use` tool with one schema for all providers.
  Actions: capture (som/vision/ax), click, double_click, right_click,
  middle_click, drag, scroll, type, key, wait, list_apps, focus_app.
- Multimodal tool-result envelope (`_multimodal=True`, OpenAI-style
  `content: [text, image_url]` parts) that flows through
  handle_function_call into the tool message. Anthropic adapter converts
  into native `tool_result` image blocks; OpenAI-compatible providers
  get the parts list directly.
- Image eviction in convert_messages_to_anthropic: only the 3 most
  recent screenshots carry real image data; older ones become text
  placeholders to cap per-turn token cost.
- Context compressor image pruning: old multimodal tool results have
  their image parts stripped instead of being skipped.
- Image-aware token estimation: each image counts as a flat 1500 tokens
  instead of its base64 char length (~1MB would have registered as
  ~250K tokens before).
- COMPUTER_USE_GUIDANCE system-prompt block — injected when the toolset
  is active.
- Session DB persistence strips base64 from multimodal tool messages.
- Trajectory saver normalises multimodal messages to text-only.
- `hermes tools` post-setup installs cua-driver via the upstream script
  and prints permission-grant instructions.
- CLI approval callback wired so destructive computer_use actions go
  through the same prompt_toolkit approval dialog as terminal commands.
- Hard safety guards at the tool level: blocked type patterns
  (curl|bash, sudo rm -rf, fork bomb), blocked key combos (empty trash,
  force delete, lock screen, log out).
- Skill `apple/macos-computer-use/SKILL.md` — universal (model-agnostic)
  workflow guide.
- Docs: `user-guide/features/computer-use.md` plus reference catalog
  entries.

44 new tests in tests/tools/test_computer_use.py covering schema
shape (universal, not Anthropic-native), dispatch routing, safety
guards, multimodal envelope, Anthropic adapter conversion, screenshot
eviction, context compressor pruning, image-aware token estimation,
run_agent helpers, and universality guarantees.

469/469 pass across tests/tools/test_computer_use.py + the affected
agent/ test suites.

- `model_tools.py` provider-gating: the tool is available to every
  provider. Providers without multi-part tool message support will see
  text-only tool results (graceful degradation via `text_summary`).
- Anthropic server-side `clear_tool_uses_20250919` — deferred;
  client-side eviction + compressor pruning cover the same cost ceiling
  without a beta header.

- macOS only. cua-driver uses private SkyLight SPIs
  (SLEventPostToPid, SLPSPostEventRecordTo,
  _AXObserverAddNotificationAndCheckRemote) that can break on any macOS
  update. Pin with HERMES_CUA_DRIVER_VERSION.
- Requires Accessibility + Screen Recording permissions — the post-setup
  prints the Settings path.

Supersedes PR #4562 (pyautogui/Quartz foreground backend, Anthropic-
native schema). Credit @0xbyt4 for the original #3816 groundwork whose
context/eviction/token design is preserved here in generic form.

agent/model_metadata.py

@@ -1455,9 +1455,79 @@ def estimate_tokens_rough(text: str) -> int:
 
 
 def estimate_messages_tokens_rough(messages: List[Dict[str, Any]]) -> int:
-    """Rough token estimate for a message list (pre-flight only)."""
-    total_chars = sum(len(str(msg)) for msg in messages)
-    return (total_chars + 3) // 4
+    """Rough token estimate for a message list (pre-flight only).
+
+    Image parts (base64 PNG/JPEG) are counted as a flat ~1500 tokens per
+    image — the Anthropic pricing model — instead of counting raw base64
+    character length. Without this, a single ~1MB screenshot would be
+    estimated at ~250K tokens and trigger premature context compression.
+    """
+    _IMAGE_TOKEN_COST = 1500
+    total_chars = 0
+    image_tokens = 0
+    for msg in messages:
+        total_chars += _estimate_message_chars(msg)
+        image_tokens += _count_image_tokens(msg, _IMAGE_TOKEN_COST)
+    return ((total_chars + 3) // 4) + image_tokens
+
+
+def _count_image_tokens(msg: Dict[str, Any], cost_per_image: int) -> int:
+    """Count image-like content parts in a message; return their token cost."""
+    count = 0
+    content = msg.get("content") if isinstance(msg, dict) else None
+    if isinstance(content, list):
+        for part in content:
+            if not isinstance(part, dict):
+                continue
+            ptype = part.get("type")
+            if ptype in ("image", "image_url", "input_image"):
+                count += 1
+    stashed = msg.get("_anthropic_content_blocks") if isinstance(msg, dict) else None
+    if isinstance(stashed, list):
+        for part in stashed:
+            if isinstance(part, dict) and part.get("type") == "image":
+                count += 1
+    # Multimodal tool results that haven't been converted yet.
+    if isinstance(content, dict) and content.get("_multimodal"):
+        inner = content.get("content")
+        if isinstance(inner, list):
+            for part in inner:
+                if isinstance(part, dict) and part.get("type") in ("image", "image_url"):
+                    count += 1
+    return count * cost_per_image
+
+
+def _estimate_message_chars(msg: Dict[str, Any]) -> int:
+    """Char count for token estimation, excluding base64 image data.
+
+    Base64 images are counted via `_count_image_tokens` instead; including
+    their raw chars here would massively overestimate token usage.
+    """
+    if not isinstance(msg, dict):
+        return len(str(msg))
+    shadow: Dict[str, Any] = {}
+    for k, v in msg.items():
+        if k == "_anthropic_content_blocks":
+            continue
+        if k == "content":
+            if isinstance(v, list):
+                cleaned = []
+                for part in v:
+                    if isinstance(part, dict):
+                        if part.get("type") in ("image", "image_url", "input_image"):
+                            cleaned.append({"type": part.get("type"), "image": "[stripped]"})
+                        else:
+                            cleaned.append(part)
+                    else:
+                        cleaned.append(part)
+                shadow[k] = cleaned
+            elif isinstance(v, dict) and v.get("_multimodal"):
+                shadow[k] = v.get("text_summary", "")
+            else:
+                shadow[k] = v
+        else:
+            shadow[k] = v
+    return len(str(shadow))
 
 
 def estimate_request_tokens_rough(
@@ -1471,13 +1541,14 @@ def estimate_request_tokens_rough(
     Includes the major payload buckets Hermes sends to providers:
     system prompt, conversation messages, and tool schemas.  With 50+
     tools enabled, schemas alone can add 20-30K tokens — a significant
-    blind spot when only counting messages.
+    blind spot when only counting messages. Image content is counted
+    at a flat per-image cost (see estimate_messages_tokens_rough).
     """
-    total_chars = 0
+    total = 0
     if system_prompt:
-        total_chars += len(system_prompt)
+        total += (len(system_prompt) + 3) // 4
     if messages:
-        total_chars += sum(len(str(msg)) for msg in messages)
+        total += estimate_messages_tokens_rough(messages)
     if tools:
-        total_chars += len(str(tools))
-    return (total_chars + 3) // 4
+        total += (len(str(tools)) + 3) // 4
+    return total