fable-engine 1.3.7__tar.gz → 1.3.9__tar.gz

{fable_engine-1.3.7/fable_engine.egg-info → fable_engine-1.3.9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: fable-engine
-Version: 1.3.7
+Version: 1.3.9
 Summary: Independent deterministic System 2 cognitive engine and mechanical time-lock MCP server
 Author: REX-codebase
 License: MIT License
@@ -133,56 +133,73 @@ red-team swarm for code, and a sealed record of what was verified.
 
 <br/>
 
-### Install
+### One package, two agent environments
+
+Fable ships as one PyPI package. The same package contains the runtime, stdio
+MCP server, and complete Agent Skill. Setup is explicit so installing an MCP
+server never silently activates workspace instructions.
+
+Run setup from the project the agent will work in:
 
 ```bash
-pip install fable-engine
+uvx --from fable-engine==1.3.9 fable-mode setup --yes
 ```
 
-Or install the MCP server in your editor:
+This resolves the pinned package in an isolated uv environment and copies the
+bundled skill to `.agents/skills/fable-mode`. Use `--dry-run` to preview or
+`--target <dir>` for another skill directory. For a persistent install, use:
 
-[![Install MCP server in VS Code](https://img.shields.io/badge/VS_Code-Install_MCP_server-007ACC?logo=visualstudiocode&logoColor=white)](vscode:mcp/install?%7B%22name%22%3A%22fable-engine%22%2C%22command%22%3A%22uvx%22%2C%22args%22%3A%5B%22--from%22%2C%22fable-engine%3D%3D1.3.7%22%2C%22fable-engine%22%5D%7D)
-[![Add to Cursor](https://img.shields.io/badge/Cursor-Add_MCP_server-black)](cursor://anysphere.cursor-deeplink/mcp/install?name=fable-engine&config=eyJjb21tYW5kIjoidXZ4IiwiYXJncyI6WyItLWZyb20iLCJmYWJsZS1lbmdpbmU9PTEuMy4zIiwiZmFibGUtZW5naW5lIl19)
+```bash
+python -m pip install fable-engine
+fable-mode setup --yes
+```
 
-These links configure Fable Engine for AI agents in VS Code Chat or Cursor. They do not install a standalone editor extension. Both use `uvx`, which downloads and runs the pinned PyPI release in an isolated environment.
+Then choose only the invocation that matches the agent environment.
 
-Point your agent at the MCP server manually:
+#### Native MCP client
+
+Run `fable-engine` as the stdio server. For example:
 
 ```jsonc
-// Claude Code: claude mcp add fable-engine -- uvx --from fable-engine==1.3.7 fable-engine
-// Cursor: ~/.cursor/mcp.json
+// Claude Code: claude mcp add fable-engine -- uvx --from fable-engine==1.3.9 fable-engine
+// Cursor or another JSON-configured client:
 {
   "mcpServers": {
     "fable-engine": {
       "command": "uvx",
-      "args": ["--from", "fable-engine==1.3.7", "fable-engine"]
+      "args": ["--from", "fable-engine==1.3.9", "fable-engine"]
     }
   }
 }
 ```
 
-Python 3.10+, zero runtime dependencies. Published on [PyPI as `fable-engine`](https://pypi.org/project/fable-engine/).
+[![Install MCP server in VS Code](https://img.shields.io/badge/VS_Code-Install_MCP_server-007ACC?logo=visualstudiocode&logoColor=white)](vscode:mcp/install?%7B%22name%22%3A%22fable-engine%22%2C%22command%22%3A%22uvx%22%2C%22args%22%3A%5B%22--from%22%2C%22fable-engine%3D%3D1.3.9%22%2C%22fable-engine%22%5D%7D)
 
-<br/>
+#### Shell sandbox with internet, no MCP host
 
-### Optional: the Agent Skill
-
-Installing `fable-engine` gives your agent the MCP tools. It does not install
-or activate the Fable Mode Agent Skill - the prompt-side workflow in
-[`skills/fable-mode`](./skills/fable-mode/SKILL.md). Nothing in the package
-activates those instructions on its own; the skill is always a separate,
-explicit opt-in.
-
-The complete skill tree ships inside the wheel. To install it into your
-project's skills directory (the cross-client `.agents/skills/` convention):
+The same package exposes a direct JSON transport. Pipe one `fable_session`
+argument object to `fable-mode call`:
 
 ```bash
-uvx --from fable-engine==1.3.7 fable-mode install-skill --yes
+printf '%s\n' '{"action":"create_session","session_name":"demo","objective":"Verify this change","time_budget_minutes":2}' \
+  | uvx --from fable-engine==1.3.9 fable-mode call
 ```
 
-This copies the skill to `.agents/skills/fable-mode`. Preview first with
-`--dry-run`, choose another location with `--target <dir>`, and reload your
-agent afterwards so it picks up the skill.
+The command uses JSON Lines: one `fable_session` argument object per input line
+and one JSON result per output line. Keep that process open for a full workflow so
+the authority timer and session stay in the same trusted runtime. A one-line pipe
+is useful for a single inspection call. Each uvx command can resolve an
+isolated environment; `pip install` is better when the sandbox keeps a Python
+environment between calls. Session data persists outside that environment in
+Fable's data directory (`FABLE_DATA_DIR` can override it).
+
+Python 3.10+, zero runtime dependencies. Published on [PyPI as `fable-engine`](https://pypi.org/project/fable-engine/).
+
+#### Skill activation remains explicit
+
+`setup` is the unified path. The older `install-skill` command remains as a
+compatible alias for skill-only installation. Neither `fable-engine` nor
+`pip install fable-engine` writes instructions into a workspace on its own.
 
 <br/>
 

{fable_engine-1.3.7 → fable_engine-1.3.9}/README.md

@@ -90,56 +90,73 @@ red-team swarm for code, and a sealed record of what was verified.
 
 <br/>
 
-### Install
+### One package, two agent environments
+
+Fable ships as one PyPI package. The same package contains the runtime, stdio
+MCP server, and complete Agent Skill. Setup is explicit so installing an MCP
+server never silently activates workspace instructions.
+
+Run setup from the project the agent will work in:
 
 ```bash
-pip install fable-engine
+uvx --from fable-engine==1.3.9 fable-mode setup --yes
 ```
 
-Or install the MCP server in your editor:
+This resolves the pinned package in an isolated uv environment and copies the
+bundled skill to `.agents/skills/fable-mode`. Use `--dry-run` to preview or
+`--target <dir>` for another skill directory. For a persistent install, use:
 
-[![Install MCP server in VS Code](https://img.shields.io/badge/VS_Code-Install_MCP_server-007ACC?logo=visualstudiocode&logoColor=white)](vscode:mcp/install?%7B%22name%22%3A%22fable-engine%22%2C%22command%22%3A%22uvx%22%2C%22args%22%3A%5B%22--from%22%2C%22fable-engine%3D%3D1.3.7%22%2C%22fable-engine%22%5D%7D)
-[![Add to Cursor](https://img.shields.io/badge/Cursor-Add_MCP_server-black)](cursor://anysphere.cursor-deeplink/mcp/install?name=fable-engine&config=eyJjb21tYW5kIjoidXZ4IiwiYXJncyI6WyItLWZyb20iLCJmYWJsZS1lbmdpbmU9PTEuMy4zIiwiZmFibGUtZW5naW5lIl19)
+```bash
+python -m pip install fable-engine
+fable-mode setup --yes
+```
 
-These links configure Fable Engine for AI agents in VS Code Chat or Cursor. They do not install a standalone editor extension. Both use `uvx`, which downloads and runs the pinned PyPI release in an isolated environment.
+Then choose only the invocation that matches the agent environment.
 
-Point your agent at the MCP server manually:
+#### Native MCP client
+
+Run `fable-engine` as the stdio server. For example:
 
 ```jsonc
-// Claude Code: claude mcp add fable-engine -- uvx --from fable-engine==1.3.7 fable-engine
-// Cursor: ~/.cursor/mcp.json
+// Claude Code: claude mcp add fable-engine -- uvx --from fable-engine==1.3.9 fable-engine
+// Cursor or another JSON-configured client:
 {
   "mcpServers": {
     "fable-engine": {
       "command": "uvx",
-      "args": ["--from", "fable-engine==1.3.7", "fable-engine"]
+      "args": ["--from", "fable-engine==1.3.9", "fable-engine"]
     }
   }
 }
 ```
 
-Python 3.10+, zero runtime dependencies. Published on [PyPI as `fable-engine`](https://pypi.org/project/fable-engine/).
+[![Install MCP server in VS Code](https://img.shields.io/badge/VS_Code-Install_MCP_server-007ACC?logo=visualstudiocode&logoColor=white)](vscode:mcp/install?%7B%22name%22%3A%22fable-engine%22%2C%22command%22%3A%22uvx%22%2C%22args%22%3A%5B%22--from%22%2C%22fable-engine%3D%3D1.3.9%22%2C%22fable-engine%22%5D%7D)
 
-<br/>
+#### Shell sandbox with internet, no MCP host
 
-### Optional: the Agent Skill
-
-Installing `fable-engine` gives your agent the MCP tools. It does not install
-or activate the Fable Mode Agent Skill - the prompt-side workflow in
-[`skills/fable-mode`](./skills/fable-mode/SKILL.md). Nothing in the package
-activates those instructions on its own; the skill is always a separate,
-explicit opt-in.
-
-The complete skill tree ships inside the wheel. To install it into your
-project's skills directory (the cross-client `.agents/skills/` convention):
+The same package exposes a direct JSON transport. Pipe one `fable_session`
+argument object to `fable-mode call`:
 
 ```bash
-uvx --from fable-engine==1.3.7 fable-mode install-skill --yes
+printf '%s\n' '{"action":"create_session","session_name":"demo","objective":"Verify this change","time_budget_minutes":2}' \
+  | uvx --from fable-engine==1.3.9 fable-mode call
 ```
 
-This copies the skill to `.agents/skills/fable-mode`. Preview first with
-`--dry-run`, choose another location with `--target <dir>`, and reload your
-agent afterwards so it picks up the skill.
+The command uses JSON Lines: one `fable_session` argument object per input line
+and one JSON result per output line. Keep that process open for a full workflow so
+the authority timer and session stay in the same trusted runtime. A one-line pipe
+is useful for a single inspection call. Each uvx command can resolve an
+isolated environment; `pip install` is better when the sandbox keeps a Python
+environment between calls. Session data persists outside that environment in
+Fable's data directory (`FABLE_DATA_DIR` can override it).
+
+Python 3.10+, zero runtime dependencies. Published on [PyPI as `fable-engine`](https://pypi.org/project/fable-engine/).
+
+#### Skill activation remains explicit
+
+`setup` is the unified path. The older `install-skill` command remains as a
+compatible alias for skill-only installation. Neither `fable-engine` nor
+`pip install fable-engine` writes instructions into a workspace on its own.
 
 <br/>
 

{fable_engine-1.3.7 → fable_engine-1.3.9}/docs/stop-ai-agents-writing-too-early.md

@@ -2,27 +2,45 @@
 
 AI coding agents often move from a plausible plan to editing files before they have gathered enough evidence. Fable Mode adds a mechanical session gate: the agent can inspect and reason, but the Fable Engine will not report write permission until the configured authority timer and proof prerequisites pass.
 
-Fable Mode is two separate pieces:
+Fable Mode ships as one package with three parts:
 
-- **Fable Engine** is the MCP server that stores session state and enforces its gates.
-- **The Fable Mode Agent Skill** is the optional workflow that tells a compatible agent how to use those gates.
+- **Fable Engine** is the stdio MCP server that stores session state and enforces its gates.
+- **The Fable Mode Agent Skill** is the bundled workflow for compatible agents.
+- **The shell transport** calls the same `fable_session` API when an agent has a shell and internet but no MCP host.
 
-Installing one does not silently activate the other.
+The package is unified; activation stays explicit. Installing or launching the
+server does not silently write Agent Skill instructions into a project.
 
-## Install the engine
+## Set up from one package
 
-You need Python 3.10+ and [uv](https://docs.astral.sh/uv/getting-started/installation/).
+You need Python 3.10+ and either [uv](https://docs.astral.sh/uv/getting-started/installation/) or pip.
 
-Run the pinned release without changing your global Python environment:
+From the project the agent will work in:
 
 ```bash
-uvx --from fable-engine==1.3.7 fable-engine
+uvx --from fable-engine==1.3.9 fable-mode setup --dry-run
+uvx --from fable-engine==1.3.9 fable-mode setup --yes
 ```
 
+`setup` copies the complete bundled skill to `.agents/skills/fable-mode`. Use
+`--target <dir>` for another skill directory. The installer refuses to
+overwrite local edits unless you pass `--force`.
+
+For a persistent environment, install the same package once:
+
+```bash
+python -m pip install fable-engine
+fable-mode setup --yes
+```
+
+## Route A: an agent with native MCP access
+
+Run `fable-engine` as its stdio MCP server.
+
 ### Claude Code
 
 ```bash
-claude mcp add fable-engine -- uvx --from fable-engine==1.3.7 fable-engine
+claude mcp add fable-engine -- uvx --from fable-engine==1.3.9 fable-engine
 ```
 
 ### Cursor or another JSON-configured MCP client
@@ -32,26 +50,33 @@ claude mcp add fable-engine -- uvx --from fable-engine==1.3.7 fable-engine
   "mcpServers": {
     "fable-engine": {
       "command": "uvx",
-      "args": ["--from", "fable-engine==1.3.7", "fable-engine"]
+      "args": ["--from", "fable-engine==1.3.9", "fable-engine"]
     }
   }
 }
 ```
 
-Restart or reload the client, then inspect its MCP tools. You should see `fable_session` plus the browser tools exposed by the server.
+Restart or reload the client. Its tool list should include `fable_session`.
 
-## Install the optional Agent Skill
+## Route B: an internet-enabled shell sandbox without MCP access
 
-From the root of the project where you want the skill:
+Pipe one `fable_session` argument object to `fable-mode call`:
 
 ```bash
-uvx --from fable-engine==1.3.7 fable-mode install-skill --dry-run
-uvx --from fable-engine==1.3.7 fable-mode install-skill --yes
+printf '%s\n' '{"action":"create_session","session_name":"parser-refactor","objective":"Refactor the parser without changing its public behavior","time_budget_minutes":2}' \
+  | uvx --from fable-engine==1.3.9 fable-mode call
 ```
 
-The default destination is `.agents/skills/fable-mode`. Use `--target <dir>` for a different skill directory. The installer refuses to overwrite local edits unless you pass `--force`.
+The command uses JSON Lines and passes each object to the same Fable handler used
+by MCP. It prints one machine-readable result per input line. Keep one process
+open for the full Think -> Prove -> Attack -> Write workflow; restarting the
+process intentionally starts a fresh authority clock. Invalid, empty, non-object,
+or over-1-MiB requests fail with a nonzero exit code and write the error to stderr.
 
-Reload the agent after installation.
+For later calls, send the same argument objects shown below. When repeated uvx
+calls are undesirable, use the persistent pip installation and replace the
+command after the pipe with `fable-mode call`. Session files live outside the
+uvx environment; set `FABLE_DATA_DIR` to choose their location.
 
 ## Run a first session
 

{fable_engine-1.3.7 → fable_engine-1.3.9}/fable_engine/actions/system3.py

@@ -282,11 +282,53 @@ def _handle_system3_evolve_paradigms(arguments: Dict[str, Any]) -> str:
     )
     pool.initialize_population(seed_paradigms=seed_paradigms)
 
-    for _ in range(generations):
-        pool.evolve_generation()
+    observed_scores = arguments.get("observed_fitness")
+    if isinstance(observed_scores, str):
+        try:
+            observed_scores = json.loads(observed_scores)
+        except Exception:
+            return "Error: 'observed_fitness' must be a JSON object keyed by genome ID or '*' default."
+
+    evolution_receipt = None
+    if observed_scores is not None:
+        if not isinstance(observed_scores, dict) or not observed_scores:
+            return "Error: 'observed_fitness' must be a non-empty object keyed by genome ID or '*' default."
+
+        def observed_evaluator(genome: CognitiveGenome) -> Dict[str, float]:
+            raw = observed_scores.get(genome.genome_id, observed_scores.get("*"))
+            if not isinstance(raw, dict):
+                raise ValueError(f"Missing observed fitness for genome '{genome.genome_id}' and no '*' default was supplied.")
+            normalized: Dict[str, float] = {}
+            for dim in pool.population[0].fitness_scores.keys() or []:
+                if dim in raw:
+                    normalized[dim] = max(0.0, min(1.0, float(raw[dim])))
+            if not normalized:
+                from fable_v2.system3 import PARETO_DIMENSIONS
+                for dim in PARETO_DIMENSIONS:
+                    if dim in raw:
+                        normalized[dim] = max(0.0, min(1.0, float(raw[dim])))
+            if not normalized:
+                raise ValueError("Observed fitness rows must contain at least one recognized Pareto dimension.")
+            return normalized
+
+        try:
+            evolution_receipt = pool.evolve_until_stable(
+                observed_evaluator,
+                max_generations=generations,
+                patience=int(arguments.get("stagnation_patience", min(4, max(1, generations)))),
+                min_improvement=float(arguments.get("min_improvement", 0.001)),
+                weights=weights,
+            )
+        except (TypeError, ValueError) as exc:
+            return f"Error: Invalid observed fitness: {exc}"
+    else:
+        for _ in range(generations):
+            pool.evolve_generation()
 
     pareto_frontier = pool.get_pareto_frontier()
-    best_genome = pool.get_best_genome(weights)
+    best_genome = pool.get_best_genome(weights) if evolution_receipt is None else max(
+        pool.population, key=lambda g: g.compute_scalar_fitness(weights)
+    )
 
     session = get_or_load_session(session_name)
     session.system3_gene_pools.append(pool.to_dict())
@@ -316,7 +358,10 @@ def _handle_system3_evolve_paradigms(arguments: Dict[str, Any]) -> str:
         f"- **Session**: `{session.session_name}`\n"
         f"- **Generations Evolved**: `{pool.generation_count}` (Population: `{len(pool.population)}`)\n"
         f"- **Rank 1 Pareto Frontier Size**: `{len(pareto_frontier)}` non-dominated solutions\n"
-        f"- **Top Archetype**: **{best_genome.paradigm_name}** (`{best_genome.genome_id}`)\n\n"
+        f"- **Top Archetype**: **{best_genome.paradigm_name}** (`{best_genome.genome_id}`)\n"
+        f"- **Measured Loop**: `{'ACTIVE - external acceptance required' if evolution_receipt else 'structural heuristic mode'}`\n"
+        + (f"- **Stop Reason**: `{evolution_receipt['stop_reason']}`\n" if evolution_receipt else "")
+        + "\n"
         f"#### 🏆 Top Rank 1 Non-Dominated Pareto Frontier:\n"
         f"| ID | Paradigm | Lat | Tput | Mem | Fault | Mod | Sec | Token | Score |\n"
         f"|---|---|---|---|---|---|---|---|---|---|\n"

{fable_engine-1.3.7 → fable_engine-1.3.9}/fable_engine/browser.py

@@ -571,18 +571,69 @@ class StealthBrowserSession:
         self.scroll_y = max(0, min(self.scroll_y + delta_y, self.document_height - self.viewport_height))
         return self._build_status()
 
+    def _ancestor_form(self, elem: DOMElement) -> Optional[DOMElement]:
+        current = elem.parent
+        while current is not None:
+            if current.tag == "form":
+                return current
+            current = current.parent
+        return None
+
+    def _submit_get_form(self, form: DOMElement) -> Dict[str, Any]:
+        """Submit a GET form through normal navigation; never synthesize a write request."""
+        method = form.attrs.get("method", "get").strip().lower() or "get"
+        if method != "get":
+            return {
+                "status": "error",
+                "error": f"Form method '{method.upper()}' is unsupported because it may mutate remote state",
+                "element_id": form.element_id,
+            }
+        action = urllib.parse.urljoin(self.url, form.attrs.get("action") or self.url)
+        pairs: List[Tuple[str, str]] = []
+        stack = list(reversed(form.children))
+        while stack:
+            node = stack.pop()
+            stack.extend(reversed(node.children))
+            name = node.attrs.get("name", "").strip()
+            if not name or node.attrs.get("disabled") is not None:
+                continue
+            if node.tag == "textarea":
+                pairs.append((name, node.attrs.get("value", node.text)))
+            elif node.tag == "input":
+                input_type = node.attrs.get("type", "text").lower()
+                if input_type == "password":
+                    return {
+                        "status": "error",
+                        "error": "GET form submission with a password field is blocked to prevent credential leakage",
+                        "element_id": form.element_id,
+                    }
+                if input_type in ("submit", "button", "reset", "file", "image"):
+                    continue
+                if input_type in ("checkbox", "radio") and "checked" not in node.attrs:
+                    continue
+                pairs.append((name, node.attrs.get("value", "on" if input_type in ("checkbox", "radio") else "")))
+        parsed = urllib.parse.urlsplit(action)
+        query = urllib.parse.parse_qsl(parsed.query, keep_blank_values=True) + pairs
+        target = urllib.parse.urlunsplit((parsed.scheme, parsed.netloc, parsed.path, urllib.parse.urlencode(query), parsed.fragment))
+        return self.open(target)
+
     def click(self, element_id: str) -> Dict[str, Any]:
-        """Follow an element's link target; non-link controls are unsupported."""
+        """Follow a link or activate a safe GET form control."""
         elem = self.elements_by_id.get(element_id)
         if not elem:
-            return {"error": f"Element '{element_id}' not found"}
+            return {"status": "error", "error": f"Element '{element_id}' not found"}
         href = elem.attrs.get("href")
         if href is not None:
             target_url = urllib.parse.urljoin(self.url, href)
             return self.open(target_url)
+        input_type = elem.attrs.get("type", "").lower()
+        if elem.tag == "button" or (elem.tag == "input" and input_type in ("submit", "image")):
+            form = self._ancestor_form(elem)
+            if form is not None:
+                return self._submit_get_form(form)
         return {
             "status": "error",
-            "error": f"Click is unsupported for <{elem.tag}> elements without a link target",
+            "error": f"Click is unsupported for <{elem.tag}> elements without a link or safe GET form target",
             "element_id": element_id,
         }
 
@@ -672,6 +723,10 @@ class StealthBrowserSession:
                     updated_value = current_value + " "
                 elif key == "Enter" and target.tag == "textarea":
                     updated_value = current_value + "\n"
+                elif key == "Enter" and target.tag == "input":
+                    form = self._ancestor_form(target)
+                    if form is not None:
+                        return self._submit_get_form(form)
                 elif len(key) == 1:
                     updated_value = current_value + key
                 if updated_value is not None and updated_value != current_value:

{fable_engine-1.3.7 → fable_engine-1.3.9}/fable_engine/fable_session.json

@@ -156,7 +156,7 @@
         "type": "number"
       },
       "curvature": {
-        "description": "Sectional curvature c > 0 of Poincar\u00e9 manifold (default 1.0).",
+        "description": "Sectional curvature c > 0 of Poincaré manifold (default 1.0).",
         "type": "number"
       },
       "d_prior": {
@@ -167,7 +167,7 @@
         ]
       },
       "dimension": {
-        "description": "Dimension of Poincar\u00e9 ball manifold (default 2).",
+        "description": "Dimension of Poincaré ball manifold (default 2).",
         "type": "integer"
       },
       "domain": {
@@ -597,6 +597,23 @@
       "user_prompt": {
         "description": "Alias for prompt: high-level design prompt describing desired web experience.",
         "type": "string"
+      },
+      "observed_fitness": {
+        "description": "Optional measured fitness object keyed by genome ID, with '*' as a default row. Enables bounded evidence-fed evolution and does not authorize deployment.",
+        "type": [
+          "object",
+          "string"
+        ]
+      },
+      "stagnation_patience": {
+        "description": "Measured generations without improvement before automatic evolution stops.",
+        "type": "integer",
+        "minimum": 1
+      },
+      "min_improvement": {
+        "description": "Minimum scalar measured-fitness gain required to reset stagnation.",
+        "type": "number",
+        "minimum": 0
       }
     },
     "required": [

{fable_engine-1.3.7 → fable_engine-1.3.9}/fable_engine/schema.py

@@ -306,6 +306,20 @@ TOOL_SCHEMA = {
                 "description": "Optional dictionary of weights across the 10 Pareto dimensions.",
                 "type": ["object", "string"]
             },
+            "observed_fitness": {
+                "description": "Optional measured fitness object keyed by genome ID, with '*' as a default row. Supplying it enables bounded evidence-fed evolution with stagnation stopping; values are clamped to [0, 1] and never authorize deployment.",
+                "type": ["object", "string"]
+            },
+            "stagnation_patience": {
+                "description": "Measured generations without improvement before automatic evolution stops (default min(4, generations)).",
+                "type": "integer",
+                "minimum": 1
+            },
+            "min_improvement": {
+                "description": "Minimum scalar measured-fitness gain required to reset stagnation (default 0.001).",
+                "type": "number",
+                "minimum": 0
+            },
             "task_complexity": {
                 "type": "number",
                 "description": "Task complexity index [0.0, 1.0] for tri-level cognitive arbitration."
@@ -625,11 +639,14 @@ BROWSER_TOOL_SCHEMAS = [
     },
     {
         "name": "browser_click",
-        "description": "Navigates to the href of a link element by its stable element ID. Elements without an href are unsupported.",
+        "title": "Activate a safe browser target",
+        "annotations": {"title": "Activate a safe browser target", "readOnlyHint": False, "destructiveHint": False, "idempotentHint": False, "openWorldHint": True},
+        "outputSchema": {"type": "object", "additionalProperties": True},
+        "description": "Activates an href-bearing link or submits a GET form through normal browser navigation. POST and other form methods are refused, and GET forms containing password fields are blocked to prevent credential leakage. Returns the resulting page status or a bounded error without bypassing existing navigation controls.",
         "inputSchema": {
             "type": "object",
             "properties": {
-                "element_id": {"type": "string", "description": "Stable element ID of an href-bearing link."},
+                "element_id": {"type": "string", "description": "Stable element ID of an href-bearing link or submit control inside a safe GET form."},
                 "session_id": {"type": "string", "description": "Optional browser tab/session identifier."}
             },
             "required": ["element_id"]

{fable_engine-1.3.7 → fable_engine-1.3.9}/fable_engine/scrapers/base.py

@@ -8,6 +8,7 @@ clamped retry/backoff, response bounds, and explicit untrusted external content
 from __future__ import annotations
 
 import datetime
+import hashlib
 import http.client
 import ipaddress
 import json
@@ -25,7 +26,7 @@ from typing import Any, Dict, List, Optional, Tuple
 
 logger = logging.getLogger("fable-engine.scrapers.base")
 
-USER_AGENT = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/122.0.0.0 Safari/537.36 FableResearch/1.3.7"
+USER_AGENT = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/122.0.0.0 Safari/537.36 FableResearch/1.3.9"
 
 MAX_RETRY_DELAY_SECONDS = 10.0
 
@@ -173,6 +174,23 @@ class DomainRateLimiter:
 GLOBAL_RATE_LIMITER = DomainRateLimiter(min_interval_seconds=0.3)
 
 
+def _canonical_origin(url: str) -> str:
+    """Return a stable, credential-free origin for provenance metadata."""
+    try:
+        parsed = urllib.parse.urlsplit(url)
+        if parsed.scheme.lower() not in ("http", "https") or not parsed.hostname:
+            return ""
+        host = parsed.hostname.lower().rstrip(".")
+        if ":" in host:
+            host = f"[{host}]"
+        port = parsed.port
+        default_port = 443 if parsed.scheme.lower() == "https" else 80
+        authority = host if port in (None, default_port) else f"{host}:{port}"
+        return f"{parsed.scheme.lower()}://{authority}"
+    except (TypeError, ValueError):
+        return ""
+
+
 @dataclass
 class ResearchResult:
     """Structured research retrieval result."""
@@ -189,6 +207,14 @@ class ResearchResult:
     def __post_init__(self):
         if "trust_level" not in self.metadata:
             self.metadata["trust_level"] = "untrusted_external_content"
+        # Every successful extraction carries deterministic integrity/provenance facts.
+        # This does not make external content trusted; it lets callers detect repeated,
+        # changed, or accidentally truncated evidence without source-specific code.
+        if self.ok:
+            encoded = self.content.encode("utf-8", errors="replace")
+            self.metadata.setdefault("content_bytes", len(encoded))
+            self.metadata.setdefault("content_sha256", hashlib.sha256(encoded).hexdigest())
+            self.metadata.setdefault("canonical_origin", _canonical_origin(self.canonical_url))
 
     def to_markdown(self) -> str:
         """Formats the structured research result as readable Markdown with explicit untrusted content boundaries."""

{fable_engine-1.3.7 → fable_engine-1.3.9}/fable_engine/server.py

@@ -237,7 +237,7 @@ def main():
                     },
                     "serverInfo": {
                         "name": "fable-engine",
-                        "version": "1.3.7"
+                        "version": "1.3.9"
                     }
                 }
             })

{fable_engine-1.3.7 → fable_engine-1.3.9}/fable_engine/session.py

@@ -1567,7 +1567,7 @@ class FableSession:
     def to_dict(self) -> Dict[str, Any]:
         """Serializes session to dictionary."""
         return {
-            "version": "1.3.7",
+            "version": "1.3.9",
             "session_name": self.session_name,
             "session_id": self.session_id,
             "objective": self.objective,