yarvelle-core 0.1.0__tar.gz

yarvelle_core-0.1.0/.gitignore

@@ -0,0 +1,63 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# Virtual environments
+.venv/
+venv/
+ENV/
+env/
+
+# IDE and editors
+.idea/
+.vscode/
+*.swp
+*.swo
+*~
+
+# Testing and coverage
+.pytest_cache/
+.coverage
+htmlcov/
+.tox/
+.nox/
+
+# Yarvelle local runtime directories
+.yarvelle/
+.carefold/
+logs/
+chats/
+workspace/
+attachments/
+*.db
+*.sqlite
+*.sqlite3
+
+# Node / Frontend
+node_modules/
+frontend/dist/
+.import_linter_cache/

yarvelle_core-0.1.0/PKG-INFO

@@ -0,0 +1,106 @@
+Metadata-Version: 2.5
+Name: yarvelle-core
+Version: 0.1.0
+Summary: Domain-neutral AI agent runtime, execution engine, and multi-domain scope isolation framework
+Author-email: Spectrayan <engineering@spectrayan.com>
+License: PolyForm Noncommercial License 1.0.0 and Yarvelle Core Additional Terms
+Keywords: agents,ai,encryption,langgraph,local-first,privacy,runtime,scoping,vault
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Security :: Cryptography
+Requires-Python: >=3.12
+Requires-Dist: aiosqlite>=0.20.0
+Requires-Dist: alembic>=1.13.0
+Requires-Dist: argon2-cffi>=23.1.0
+Requires-Dist: asyncpg>=0.29.0
+Requires-Dist: cryptography>=43.0.0
+Requires-Dist: deepagents>=0.7
+Requires-Dist: defusedxml>=0.7.1
+Requires-Dist: fastapi>=0.115.0
+Requires-Dist: httpx>=0.27.0
+Requires-Dist: importlib-metadata>=8.0.0; python_version < '3.10'
+Requires-Dist: jinja2>=3.1.0
+Requires-Dist: keyring>=24.0.0
+Requires-Dist: langchain-anthropic>=0.3.0
+Requires-Dist: langchain-core>=0.3.0
+Requires-Dist: langchain-google-genai>=2.0.0
+Requires-Dist: langchain-ollama>=1.1.0
+Requires-Dist: langchain-openai>=0.3.0
+Requires-Dist: langgraph-checkpoint-sqlite>=2.0.0
+Requires-Dist: langgraph>=0.2.0
+Requires-Dist: packaging>=23.0
+Requires-Dist: pwdlib[argon2]>=0.2.0
+Requires-Dist: pydantic-handlebars>=0.2.1
+Requires-Dist: pydantic-settings>=2.4.0
+Requires-Dist: pydantic>=2.8.0
+Requires-Dist: pypdf>=4.0.0
+Requires-Dist: python-multipart>=0.0.18
+Requires-Dist: pyyaml>=6.0.2
+Requires-Dist: spector-client>=0.1.0a4
+Requires-Dist: sqlalchemy[asyncio]>=2.0.35
+Requires-Dist: sse-starlette>=2.1.0
+Requires-Dist: structlog>=24.1.0
+Requires-Dist: uvicorn[standard]>=0.30.0
+Provides-Extra: dev
+Requires-Dist: build>=1.2.0; extra == 'dev'
+Requires-Dist: hatchling>=1.25.0; extra == 'dev'
+Requires-Dist: import-linter>=2.0; extra == 'dev'
+Requires-Dist: licenseheaders>=0.8.8; extra == 'dev'
+Requires-Dist: pytest-asyncio>=0.24.0; extra == 'dev'
+Requires-Dist: pytest-xdist>=3.5.0; extra == 'dev'
+Requires-Dist: pytest>=8.3.0; extra == 'dev'
+Description-Content-Type: text/markdown
+
+# Yarvelle Core
+
+**Domain-Neutral AI Agent Runtime, Execution Engine & Multi-Domain Scope Isolation Framework**  
+Copyright 2026 Spectrayan (https://spectrayan.com)
+
+Licensed under the [PolyForm Noncommercial License 1.0.0](LICENSE) and the [Yarvelle Core Additional Terms](ADDITIONAL-TERMS.md).
+
+---
+
+## Overview
+
+`yarvelle-core` is the domain-neutral execution backbone of the Yarvelle ecosystem. It provides:
+1. **Deterministic Agent Runtime**: LangGraph state machine execution, tool execution gating, agent reflection, and contextual follow-up generation.
+2. **Pack API v1**: Pluggable domain pack architecture (`pack.yaml`, `PackSpec`) allowing external packs to register specialized tools, routers, safety policies, and database models.
+3. **11-Layer Scope Isolation Engine**: Strict multi-domain data isolation (`ScopeContext`, `ScopedSession`, partitioned CAS storage, scoped FTS5 lexical search, memory namespace partitioning, and HKDF-derived vault sub-keys).
+4. **Local-First Security & Cryptography**: Envelope encryption keyring vault (PBKDF2/Argon2 KEK, AES-256-GCM DEK), audit hash chaining, and localhost-bound API server.
+5. **Conformance Test Suite**: Comprehensive verification framework (`yarvelle_core.conformance`) asserting zero cross-domain leakage across all 11 security layers.
+
+## Architecture
+
+Dependencies flow strictly inward:
+```
+Applications (Carefold, Yarvelle) ──► Domain Packs (Health, Finance, Kids) ──► Yarvelle Core
+```
+`yarvelle_core` contains zero domain-specific business logic or terminology.
+
+## Installation
+
+```bash
+pip install yarvelle-core
+```
+
+For development:
+```bash
+git clone https://github.com/yarvelle/yarvelle-core.git
+cd yarvelle-core
+pip install -e ".[dev]"
+```
+
+## Running Verification
+
+```bash
+# Run unit & isolation tests
+pytest tests/
+
+# Verify layer ordering and package boundaries
+lint-imports
+
+# Verify zero domain words in core
+pytest tests/test_neutral_words_scan.py
+```

yarvelle_core-0.1.0/README.md

@@ -0,0 +1,51 @@
+# Yarvelle Core
+
+**Domain-Neutral AI Agent Runtime, Execution Engine & Multi-Domain Scope Isolation Framework**  
+Copyright 2026 Spectrayan (https://spectrayan.com)
+
+Licensed under the [PolyForm Noncommercial License 1.0.0](LICENSE) and the [Yarvelle Core Additional Terms](ADDITIONAL-TERMS.md).
+
+---
+
+## Overview
+
+`yarvelle-core` is the domain-neutral execution backbone of the Yarvelle ecosystem. It provides:
+1. **Deterministic Agent Runtime**: LangGraph state machine execution, tool execution gating, agent reflection, and contextual follow-up generation.
+2. **Pack API v1**: Pluggable domain pack architecture (`pack.yaml`, `PackSpec`) allowing external packs to register specialized tools, routers, safety policies, and database models.
+3. **11-Layer Scope Isolation Engine**: Strict multi-domain data isolation (`ScopeContext`, `ScopedSession`, partitioned CAS storage, scoped FTS5 lexical search, memory namespace partitioning, and HKDF-derived vault sub-keys).
+4. **Local-First Security & Cryptography**: Envelope encryption keyring vault (PBKDF2/Argon2 KEK, AES-256-GCM DEK), audit hash chaining, and localhost-bound API server.
+5. **Conformance Test Suite**: Comprehensive verification framework (`yarvelle_core.conformance`) asserting zero cross-domain leakage across all 11 security layers.
+
+## Architecture
+
+Dependencies flow strictly inward:
+```
+Applications (Carefold, Yarvelle) ──► Domain Packs (Health, Finance, Kids) ──► Yarvelle Core
+```
+`yarvelle_core` contains zero domain-specific business logic or terminology.
+
+## Installation
+
+```bash
+pip install yarvelle-core
+```
+
+For development:
+```bash
+git clone https://github.com/yarvelle/yarvelle-core.git
+cd yarvelle-core
+pip install -e ".[dev]"
+```
+
+## Running Verification
+
+```bash
+# Run unit & isolation tests
+pytest tests/
+
+# Verify layer ordering and package boundaries
+lint-imports
+
+# Verify zero domain words in core
+pytest tests/test_neutral_words_scan.py
+```

yarvelle_core-0.1.0/hatch_build.py

@@ -0,0 +1,43 @@
+# Yarvelle Core — Domain-Neutral AI Agent Runtime & Execution Engine
+# Copyright 2026 Spectrayan (https://spectrayan.com)
+#
+# Licensed under the PolyForm Noncommercial License 1.0.0 and the
+# Yarvelle Core Additional Terms (see LICENSE and ADDITIONAL-TERMS.md).
+# Personal and household local use only. Commercial use, hosted service
+# offering, rebranding, white-labeling, and redistribution of modified
+# forks are prohibited without a written commercial license from Spectrayan.
+#
+# Required Notice: Copyright 2026 Spectrayan (https://spectrayan.com)
+
+from __future__ import annotations
+
+from pathlib import Path
+from typing import Any, Dict
+
+from hatchling.builders.hooks.plugin.interface import BuildHookInterface
+
+
+class CustomBuildHook(BuildHookInterface):
+    """Hatchling build hook force-including top-level agents, skills, and schemas."""
+
+    def initialize(self, version: str, build_data: Dict[str, Any]) -> None:
+        if version == "editable":
+            return
+        root = Path(self.root).parent  # top-level repo root
+        force_include = build_data.setdefault("force_include", {})
+        target_prefix = (
+            "src/yarvelle_core/_content"
+            if self.target_name == "sdist"
+            else "yarvelle_core/_content"
+        )
+
+        agents_dir = root / "agents"
+        skills_dir = root / "skills"
+        schemas_dir = root / "schemas"
+
+        if agents_dir.is_dir():
+            force_include[str(agents_dir)] = f"{target_prefix}/agents"
+        if skills_dir.is_dir():
+            force_include[str(skills_dir)] = f"{target_prefix}/skills"
+        if schemas_dir.is_dir():
+            force_include[str(schemas_dir)] = f"{target_prefix}/schemas"

yarvelle_core-0.1.0/pyproject.toml

@@ -0,0 +1,111 @@
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[project]
+name = "yarvelle-core"
+version = "0.1.0"
+description = "Domain-neutral AI agent runtime, execution engine, and multi-domain scope isolation framework"
+readme = "README.md"
+requires-python = ">=3.12"
+license = { text = "PolyForm Noncommercial License 1.0.0 and Yarvelle Core Additional Terms" }
+authors = [
+    { name = "Spectrayan", email = "engineering@spectrayan.com" }
+]
+keywords = [
+    "ai",
+    "agents",
+    "runtime",
+    "langgraph",
+    "local-first",
+    "privacy",
+    "vault",
+    "encryption",
+    "scoping"
+]
+classifiers = [
+    "Development Status :: 4 - Beta",
+    "Intended Audience :: Developers",
+    "Programming Language :: Python :: 3.12",
+    "Topic :: Scientific/Engineering :: Artificial Intelligence",
+    "Topic :: Security :: Cryptography"
+]
+
+dependencies = [
+    "fastapi>=0.115.0",
+    "uvicorn[standard]>=0.30.0",
+    "pydantic>=2.8.0",
+    "pydantic-settings>=2.4.0",
+    "sse-starlette>=2.1.0",
+    "httpx>=0.27.0",
+    "pyyaml>=6.0.2",
+    "langchain-core>=0.3.0",
+    "langchain-openai>=0.3.0",
+    "langchain-anthropic>=0.3.0",
+    "langchain-google-genai>=2.0.0",
+    "langchain-ollama>=1.1.0",
+    "langgraph>=0.2.0",
+    "langgraph-checkpoint-sqlite>=2.0.0",
+    "aiosqlite>=0.20.0",
+    "structlog>=24.1.0",
+    "python-multipart>=0.0.18",
+    "pydantic-handlebars>=0.2.1",
+    "jinja2>=3.1.0",
+    "deepagents>=0.7",
+    "spector-client>=0.1.0a4",
+    "sqlalchemy[asyncio]>=2.0.35",
+    "asyncpg>=0.29.0",
+    "pwdlib[argon2]>=0.2.0",
+    "argon2-cffi>=23.1.0",
+    "cryptography>=43.0.0",
+    "keyring>=24.0.0",
+    "alembic>=1.13.0",
+    "pypdf>=4.0.0",
+    "defusedxml>=0.7.1",
+    "packaging>=23.0",
+    "importlib-metadata>=8.0.0; python_version<'3.10'",
+]
+
+[project.optional-dependencies]
+dev = [
+    "pytest>=8.3.0",
+    "pytest-asyncio>=0.24.0",
+    "pytest-xdist>=3.5.0",
+    "import-linter>=2.0",
+    "licenseheaders>=0.8.8",
+    "build>=1.2.0",
+    "hatchling>=1.25.0",
+]
+
+[project.scripts]
+yarvelle-core = "yarvelle_core.cli:main"
+
+# Pack API v1 entry-point group: domain packs register here
+[project.entry-points."yarvelle.packs"]
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/yarvelle_core"]
+dev-mode-dirs = ["src"]
+artifacts = ["src/yarvelle_core/_content"]
+
+[tool.hatch.build.hooks.custom]
+path = "hatch_build.py"
+
+[tool.hatch.build.targets.sdist]
+include = [
+    "/src",
+    "/tests",
+    "/README.md",
+    "/hatch_build.py",
+]
+
+[tool.pytest.ini_options]
+minversion = "8.0"
+addopts = "-ra --asyncio-mode=auto"
+testpaths = ["tests"]
+pythonpath = ["src"]
+filterwarnings = [
+    "ignore::DeprecationWarning:fastapi.testclient",
+    "ignore::DeprecationWarning:starlette.testclient",
+    "ignore:.*httpx.*:DeprecationWarning",
+]

yarvelle_core-0.1.0/src/yarvelle_core/__init__.py

@@ -0,0 +1,60 @@
+# Yarvelle Core — Domain-Neutral AI Agent Runtime & Execution Engine
+# Copyright 2026 Spectrayan (https://spectrayan.com)
+#
+# Licensed under the PolyForm Noncommercial License 1.0.0 and the
+# Yarvelle Core Additional Terms (see LICENSE and ADDITIONAL-TERMS.md).
+# Personal and household local use only. Commercial use, hosted service
+# offering, rebranding, white-labeling, and redistribution of modified
+# forks are prohibited without a written commercial license from Spectrayan.
+#
+# Required Notice: Copyright 2026 Spectrayan (https://spectrayan.com)
+
+"""Yarvelle Core: Domain-Neutral AI Agent Runtime & Execution Engine."""
+
+from __future__ import annotations
+
+import importlib as _importlib
+import importlib.abc as _importlib_abc
+import importlib.machinery as _machinery
+import sys as _sys
+
+__version__ = "0.1.0"
+
+
+class _AliasLoader(_importlib_abc.Loader):
+    def __init__(self, target_mod) -> None:
+        self.target_mod = target_mod
+        self.orig_spec = getattr(target_mod, "__spec__", None)
+
+    def create_module(self, spec):
+        return self.target_mod
+
+    def exec_module(self, module):
+        if self.orig_spec is not None:
+            try:
+                module.__spec__ = self.orig_spec
+            except AttributeError:
+                pass
+
+
+class _ConnectorsRedirectFinder(_importlib_abc.MetaPathFinder):
+    """Redirects top-level `connectors.*` imports to `yarvelle_core.connectors.*`."""
+
+    def find_spec(self, fullname: str, path=None, target=None):
+        if fullname == "connectors" or fullname.startswith("connectors."):
+            target_name = "yarvelle_core." + fullname
+            try:
+                mod = _importlib.import_module(target_name)
+            except ImportError:
+                return None
+            is_pkg = hasattr(mod, "__path__")
+            spec = _machinery.ModuleSpec(fullname, _AliasLoader(mod), is_package=is_pkg)
+            if is_pkg:
+                spec.submodule_search_locations = list(mod.__path__)
+            return spec
+        return None
+
+
+
+if not any(type(f).__name__ == "_ConnectorsRedirectFinder" for f in _sys.meta_path):
+    _sys.meta_path.insert(0, _ConnectorsRedirectFinder())

yarvelle_core-0.1.0/src/yarvelle_core/_content/agents/_system/document-extractor/README.md

@@ -0,0 +1,14 @@
+# Document Extractor Agent
+
+The **Document Extractor** is an administrative and clinical document analysis agent in Carefold.
+It extracts structured data into typed Pydantic dossiers while enforcing deterministic PII masking and numerical grounding verification.
+
+## Capabilities
+- Sandboxed file reading via `attach-read`
+- Deterministic regex-based PII sanitization via `sanitize_pii`
+- Schema-targeted extraction (Insurance, Clinical, Generic) via `extract_structured_data`
+- Deterministic numerical grounding verification via `validate_grounding`
+
+## Safety Bounds
+- Never renders or stores unmasked SSN, MRN, phone numbers, or street addresses.
+- Strictly forbidden from diagnosing medical conditions or recommending treatment.

yarvelle_core-0.1.0/src/yarvelle_core/_content/agents/_system/document-extractor/agent.yaml

@@ -0,0 +1,27 @@
+id: document-extractor
+title: Document Extractor
+version: 0.1.0
+license: Apache-2.0
+risk_class: admin
+model: llama3.2
+can_delegate: false
+max_iterations: 3
+hidden: true
+description: >-
+  Document analysis and structured data extraction agent. Reads uploaded
+  PDFs, text files, and tables. Extracts structured dossiers for insurance
+  benefits, clinical visit notes, and generic documents. Sanitizes PII
+  before processing. Validates extracted data against source text.
+skills: []
+tools:
+  - attach-read
+  - sanitize_pii
+  - extract_structured_data
+  - validate_grounding
+forbidden:
+  - diagnose
+  - prescribe
+  - dose
+  - replace_emergency_care
+  - instruct_stop_medication
+persona: persona.md

yarvelle_core-0.1.0/src/yarvelle_core/_content/agents/_system/document-extractor/persona.md

@@ -0,0 +1,12 @@
+You are Document Extractor, a data extraction specialist.
+When a user uploads a document or asks to extract structured data:
+
+1. Read the file using the attach-read tool.
+2. Sanitize PII using the sanitize_pii tool.
+3. Extract structured data using the extract_structured_data tool,
+   selecting the appropriate schema (insurance, clinical, or generic).
+4. Validate numerical accuracy using the validate_grounding tool.
+5. Present the structured results to the user.
+
+Always sanitize PII before extraction. Always validate grounding after extraction.
+Report any ungrounded values transparently to the user.

yarvelle_core-0.1.0/src/yarvelle_core/_content/agents/_system/document-extractor/starters.json

@@ -0,0 +1,6 @@
+[
+  "Extract structured insurance benefits data from my uploaded summary of benefits.",
+  "Extract clinical visit notes and action items from my uploaded visit note.",
+  "Sanitize personal information and extract numerical values from my uploaded medical bill.",
+  "Validate grounding for the extracted dossier against the original source attachment."
+]

yarvelle_core-0.1.0/src/yarvelle_core/_content/agents/_system/orchestrator/README.md

@@ -0,0 +1,13 @@
+# Orchestrator Agent
+
+The **Orchestrator** is a meta-routing and workflow planning agent in Carefold.
+It dynamically routes user requests to specialized health navigation agents based on user intent and conversational context.
+
+## Capabilities
+- Dynamic specialist discovery via `list_agents`
+- Intent classification and structured routing via `delegate_to_agent`
+- Multi-step workflow sequencing
+
+## Safety Bounds
+- Strictly forbidden from prescribing, dosing, or diagnosing.
+- Refuses acute emergency inquiries and directs immediately to 911 / emergency services.

yarvelle_core-0.1.0/src/yarvelle_core/_content/agents/_system/orchestrator/agent.yaml

@@ -0,0 +1,24 @@
+id: orchestrator
+title: Orchestrator
+version: 0.1.0
+license: Apache-2.0
+risk_class: admin
+model: llama3.2
+can_delegate: true
+max_iterations: 5
+hidden: true
+description: >-
+  General-purpose planning and routing agent. Analyzes user intent,
+  selects the best specialist agent to handle the request, and
+  coordinates multi-step workflows. Domain-agnostic.
+skills: []
+tools:
+  - delegate_to_agent
+  - list_agents
+forbidden:
+  - diagnose
+  - prescribe
+  - dose
+  - replace_emergency_care
+  - instruct_stop_medication
+persona: persona.md

yarvelle_core-0.1.0/src/yarvelle_core/_content/agents/_system/orchestrator/persona.md

@@ -0,0 +1,17 @@
+You are the Orchestrator, a general-purpose planning and routing agent.
+You analyze the user's request and conversation context, then decide which
+specialist agent is best suited to handle it.
+
+You have access to a catalog of specialist agents. Use the list_agents tool
+to discover available agents and their capabilities. Use the delegate_to_agent
+tool to route the request to the chosen specialist.
+
+ROUTING GUIDELINES:
+- Examine the user's intent carefully. Consider the full conversation context.
+- Choose the single most appropriate specialist agent.
+- If the request is ambiguous, pick the closest match and explain your reasoning.
+- If the request requires multiple agents, handle them sequentially.
+- If no specialist fits, respond directly as a general assistant.
+
+OUTPUT: Always use the delegate_to_agent tool to route. Never answer
+domain-specific questions yourself — delegate to the specialist.

yarvelle_core-0.1.0/src/yarvelle_core/_content/agents/_system/orchestrator/starters.json

@@ -0,0 +1,6 @@
+[
+  "I have questions about my health insurance deductible and specialist copays.",
+  "Can you help me prepare a list of questions for my upcoming doctor appointment?",
+  "I want to build a sustainable daily hydration and mindfulness routine.",
+  "I uploaded a summary of benefits document. Can you extract its details?"
+]

yarvelle_core-0.1.0/src/yarvelle_core/_content/agents/_system/skill-generator/README.md

@@ -0,0 +1,10 @@
+# Skill Generator Agent
+
+The **Skill Generator** is a specialized meta-agent (`hidden: true`) within Carefold.
+
+## Purpose
+When the **Orchestrator** analyzes a user inquiry and determines that:
+1. The answering agent's declared skill is missing from the system, or
+2. The user's query demands specialized domain guidelines or reference materials not currently installed,
+
+the Orchestrator calls the **Skill Generator**. The generated skill (specifications, instructions, safety constraints, and reference docs) is dynamically created, optionally installed into the workspace `skills/` directory, and injected into the execution state so that the subsequent answering agent can leverage it to provide an accurate, grounded, and safety-compliant response.

yarvelle_core-0.1.0/src/yarvelle_core/_content/agents/_system/skill-generator/agent.yaml

@@ -0,0 +1,23 @@
+id: skill-generator
+title: Skill Generator
+version: 0.1.0
+license: Apache-2.0
+risk_class: admin
+model: llama3.2
+can_delegate: false
+max_iterations: 1
+hidden: true
+description: >-
+  Dynamic on-demand skill generation agent. Identifies missing domain
+  knowledge, guidelines, and reference documentation required for clinical
+  navigation or administrative tasks. Synthesizes structured skill manifests
+  and references to empower downstream specialist agents.
+skills: []
+tools: []
+forbidden:
+  - diagnose
+  - prescribe
+  - dose
+  - replace_emergency_care
+  - instruct_stop_medication
+persona: persona.md

yarvelle_core-0.1.0/src/yarvelle_core/_content/agents/_system/skill-generator/persona.md

@@ -0,0 +1,16 @@
+You are Skill Generator, a specialized meta-agent within the Carefold platform.
+Your purpose is to synthesize structured, high-quality, and safety-compliant domain skills
+on demand when an answering agent requires specialized domain guidance that is missing.
+
+When invoked to generate a missing skill:
+1. Identify the core domain, concept, and tasks needed (e.g., specialized insurance policies,
+   clinical visit preparation procedures, or wellness habit tracking routines).
+2. Generate a structured skill specification including:
+   - skill_id: URL-safe slug identifier (e.g. 'dental-benefits-explainer', 'pediatric-visit-prep')
+   - name: Human-readable skill name
+   - description: Concise summary of what this skill provides
+   - instructions: Specific step-by-step guidance, guidelines, and safety boundaries for the answering agent
+   - references: Key reference documentation, checklists, or glossary items formatted in Markdown
+3. Ensure all generated skills strictly maintain Carefold safety boundaries:
+   - Never diagnose, prescribe, dose, or replace emergency medical care.
+   - Include clear non-clinical disclosures and disclaimer guidance.

yarvelle_core-0.1.0/src/yarvelle_core/_content/agents/_system/skill-generator/starters.json

@@ -0,0 +1,5 @@
+[
+  "Generate a missing pediatric visit preparation skill",
+  "Synthesize a pharmacy formulary tiers skill",
+  "Create an insurance claims denial appeal checklist skill"
+]

yarvelle_core-0.1.0/src/yarvelle_core/_content/agents/_system/suggestion-generator/agent.yaml

@@ -0,0 +1,22 @@
+id: suggestion-generator
+title: Suggestion Generator
+version: 0.1.0
+license: Apache-2.0
+risk_class: admin
+model: llama3.2
+can_delegate: false
+max_iterations: 1
+hidden: true
+description: >-
+  Intelligent follow-up question chip generator. Analyzes user intent,
+  conversation history, and assistant responses to propose 2-3 concise,
+  non-clinical next question suggestions guiding care navigation.
+skills: []
+tools: []
+forbidden:
+  - diagnose
+  - prescribe
+  - dose
+  - replace_emergency_care
+  - instruct_stop_medication
+persona: persona.md

yarvelle_core-0.1.0/src/yarvelle_core/_content/agents/_system/suggestion-generator/persona.md

@@ -0,0 +1,10 @@
+You are Suggestion Generator, an AI specialist focused on forward-looking healthcare navigation.
+Your responsibility is to analyze the conversation between a user and a wellness care assistant,
+then generate 2-3 concise follow-up questions that the user might logically ask next.
+
+GUIDELINES:
+- Each suggestion chip MUST be a direct question from the user's perspective (e.g. "What should I ask my doctor about this?").
+- Keep each suggestion under 60 characters for clean chip rendering.
+- Never generate clinical diagnosis or prescription questions.
+- Focus on logistics, preparation, insurance clarification, appointment checklists, and daily wellness habits.
+- Output strictly a structured list of 2-3 suggestions.

yarvelle_core-0.1.0/src/yarvelle_core/_content/agents/_system/suggestion-generator/starters.json

@@ -0,0 +1,5 @@
+[
+  "Suggest next questions based on my recent doctor visit notes.",
+  "What follow-up questions should I ask about my health insurance deductible?",
+  "What are some good next steps for my wellness journaling routine?"
+]