diff --git a/examples/document-review.yaml b/examples/document-review.yaml
deleted file mode 100644
index 677f195..0000000
--- a/examples/document-review.yaml
+++ /dev/null
@@ -1,498 +0,0 @@
-# Document Review & Refinement Workflow
-#
-# This example demonstrates a multi-agent, multi-provider workflow that
-# reviews and refines a markdown document for clarity, readability, and
-# accuracy. It shows:
-# - Parallel review agents specializing in different quality dimensions
-# - Multi-provider usage (Claude for analysis, Copilot for refinement)
-# - MCP server integration (web-search, context7, ms-learn)
-# - Human-in-the-loop gate for clarification and approval
-# - Iterative refinement loop with quality scoring
-# - Context passing between agents with explicit mode
-#
-# Usage:
-#   conductor run examples/document-review.yaml \
-#     --input document="$(cat my-document.md)" \
-#     --input audience="developers"
-#
-# Requirements:
-#   - GitHub Copilot CLI configured (for copilot agents)
-#   - ANTHROPIC_API_KEY set (for claude agents)
-#   - Node.js installed (for MCP servers)
-
-workflow:
-  name: document-review
-  description: >
-    Review and refine a markdown document using parallel specialist agents
-    across multiple providers, with human-in-the-loop approval.
-  version: "1.0.0"
-  entry_point: document_analyzer
-
-  input:
-    document:
-      type: string
-      required: true
-      description: The markdown document content to review and refine
-    audience:
-      type: string
-      required: false
-      default: "general"
-      description: Target audience (e.g., developers, executives, general)
-
-  runtime:
-    provider: copilot
-    default_model: claude-opus-4.6
-    mcp_servers:
-      web-search:
-        command: npx
-        args: ["-y", "open-websearch@latest"]
-        env:
-          MODE: stdio
-          DEFAULT_SEARCH_ENGINE: duckduckgo
-          ALLOWED_SEARCH_ENGINES: duckduckgo,brave,exa
-        tools: ["*"]
-      context7:
-        command: npx
-        args: ["-y", "@upstash/context7-mcp@latest"]
-        tools: ["*"]
-      ms-learn:
-        type: http
-        url: https://learn.microsoft.com/api/mcp
-        tools: ["*"]
-      work-iq:
-        command: npx
-        args: [ "-y", "@microsoft/workiq", "mcp" ]
-        tools: ["*"]
-
-  context:
-    mode: explicit
-
-  limits:
-    max_iterations: 100
-
-  cost:
-    show_per_agent: true
-    show_summary: true
-
-agents:
-  # ── Step 1: Analyze the document structure and content ──────────────
-  - name: document_analyzer
-    model: claude-opus-4.6
-    description: Analyzes the document structure, identifies topics, and flags areas for review
-    input:
-      - workflow.input.document
-      - workflow.input.audience
-    tools: []
-    prompt: |
-      You are a senior technical editor. Analyze the following markdown document
-      and provide a structured assessment to guide parallel reviewers.
-
-      **Target Audience:** {{ workflow.input.audience }}
-
-      **Document:**
-      {{ workflow.input.document }}
-
-      Provide:
-      1. A brief summary of the document's purpose and scope
-      2. An outline of the document structure (headings, sections)
-      3. A list of key topics and technical claims that need fact-checking
-      4. Initial observations about tone, formatting, and organization
-      5. Any terms, acronyms, or references that may need clarification
-    output:
-      summary:
-        type: string
-        description: Brief summary of the document's purpose
-      structure:
-        type: array
-        description: Document structure outline (sections/headings)
-      claims_to_verify:
-        type: array
-        description: Technical claims and facts that need verification
-      initial_observations:
-        type: string
-        description: Observations about tone, formatting, organization
-      terms_to_clarify:
-        type: array
-        description: Terms, acronyms, or references needing clarification
-    routes:
-      - to: parallel_reviewers
-
-  # ── Step 2: Parallel specialist reviewers ───────────────────────────
-
-  - name: clarity_reviewer
-    model: claude-opus-4.6
-    description: Reviews the document for clarity, logical flow, and structure
-    input:
-      - workflow.input.document
-      - workflow.input.audience
-      - document_analyzer.output
-    tools: []
-    prompt: |
-      You are a clarity and structure specialist. Review this document for
-      logical flow, clear communication, and effective structure.
-
-      **Target Audience:** {{ workflow.input.audience }}
-      **Document Purpose:** {{ document_analyzer.output.summary }}
-
-      **Document:**
-      {{ workflow.input.document }}
-
-      **Structure Analysis:**
-      {{ document_analyzer.output.structure | json }}
-
-      Evaluate and provide specific feedback on:
-      1. **Logical Flow** — Do sections follow a logical progression?
-      2. **Clarity** — Are ideas expressed clearly and unambiguously?
-      3. **Structure** — Are headings, paragraphs, and lists used effectively?
-      4. **Transitions** — Do sections connect smoothly?
-      5. **Completeness** — Are there gaps or missing explanations?
-
-      For each issue found, provide:
-      - The specific text or section with the problem
-      - What the problem is
-      - A suggested improvement
-    output:
-      score:
-        type: number
-        description: Clarity score from 1-10
-      issues:
-        type: array
-        description: List of clarity issues with suggested fixes
-      strengths:
-        type: array
-        description: What the document does well in terms of clarity
-      recommendations:
-        type: string
-        description: Overall recommendations for clarity improvements
-
-  - name: readability_reviewer
-    description: Reviews the document for readability, tone, and audience fit
-    model: claude-sonnet-4.5
-    input:
-      - workflow.input.document
-      - workflow.input.audience
-      - document_analyzer.output.summary
-    tools: []
-    prompt: |
-      You are a readability and tone specialist. Review this document for
-      readability, appropriate tone, and audience suitability.
-
-      **Target Audience:** {{ workflow.input.audience }}
-      **Document Purpose:** {{ document_analyzer.output.summary }}
-
-      **Document:**
-      {{ workflow.input.document }}
-
-      Evaluate and provide specific feedback on:
-      1. **Readability** — Is the language accessible to the target audience?
-      2. **Tone** — Is the tone consistent and appropriate?
-      3. **Sentence Structure** — Are sentences concise and varied?
-      4. **Jargon** — Is technical language used appropriately for the audience?
-      5. **Formatting** — Are markdown features (headings, lists, code blocks,
-         bold/italic) used effectively for scannability?
-
-      For each issue found, provide:
-      - The specific text or section
-      - What the problem is
-      - A rewritten version
-    output:
-      score:
-        type: number
-        description: Readability score from 1-10
-      issues:
-        type: array
-        description: List of readability issues with rewrites
-      tone_assessment:
-        type: string
-        description: Assessment of the document's tone
-      recommendations:
-        type: string
-        description: Overall recommendations for readability improvements
-
-  - name: accuracy_checker
-    model: claude-opus-4.6
-    description: Verifies factual claims, technical accuracy, and references using web search and documentation
-    input:
-      - workflow.input.document
-      - document_analyzer.output.claims_to_verify
-      - document_analyzer.output.terms_to_clarify
-    tools: null
-    system_prompt: |
-      You are a fact-checker and technical accuracy reviewer. You have access to
-      web search, Context7 documentation lookup, and Microsoft Learn tools.
-      Use these tools to verify claims, check technical details, and validate
-      references found in the document.
-    prompt: |
-      Verify the factual accuracy and technical correctness of this document.
-
-      **Document:**
-      {{ workflow.input.document }}
-
-      **Claims to Verify:**
-      {% for claim in document_analyzer.output.claims_to_verify %}
-      - {{ claim }}
-      {% endfor %}
-
-      **Terms to Clarify:**
-      {% for term in document_analyzer.output.terms_to_clarify %}
-      - {{ term }}
-      {% endfor %}
-
-      For each claim or technical statement:
-      1. Use your tools to verify the accuracy
-      2. Check if referenced APIs, libraries, or tools are current
-      3. Validate any version numbers, URLs, or code examples
-      4. Look up official documentation where relevant
-
-      Provide:
-      - A list of verified facts (confirmed accurate)
-      - A list of inaccuracies with corrections and sources
-      - A list of outdated information with current alternatives
-      - Any claims that could not be verified
-    output:
-      score:
-        type: number
-        description: Accuracy score from 1-10
-      verified_facts:
-        type: array
-        description: Claims confirmed as accurate
-      inaccuracies:
-        type: array
-        description: Incorrect claims with corrections and sources
-      outdated_info:
-        type: array
-        description: Outdated information with current alternatives
-      unverifiable:
-        type: array
-        description: Claims that could not be verified
-      recommendations:
-        type: string
-        description: Overall accuracy recommendations
-
-  # ── Step 3: Synthesize all reviewer feedback ────────────────────────
-  - name: review_synthesizer
-    model: claude-opus-4.6
-    description: Consolidates feedback from all reviewers into a prioritized action plan
-    input:
-      - workflow.input.audience
-      - document_analyzer.output.summary
-      - parallel_reviewers.outputs
-    tools: []
-    prompt: |
-      You are a senior editor synthesizing feedback from three specialist
-      reviewers. Consolidate their findings into a clear, prioritized action plan.
-
-      **Document Purpose:** {{ document_analyzer.output.summary }}
-      **Target Audience:** {{ workflow.input.audience }}
-
-      **Clarity Review (Score: {{ parallel_reviewers.outputs.clarity_reviewer.score }}/10):**
-      Issues: {{ parallel_reviewers.outputs.clarity_reviewer.issues | json }}
-      Recommendations: {{ parallel_reviewers.outputs.clarity_reviewer.recommendations }}
-
-      **Readability Review (Score: {{ parallel_reviewers.outputs.readability_reviewer.score }}/10):**
-      Issues: {{ parallel_reviewers.outputs.readability_reviewer.issues | json }}
-      Tone: {{ parallel_reviewers.outputs.readability_reviewer.tone_assessment }}
-      Recommendations: {{ parallel_reviewers.outputs.readability_reviewer.recommendations }}
-
-      **Accuracy Review (Score: {{ parallel_reviewers.outputs.accuracy_checker.score }}/10):**
-      Inaccuracies: {{ parallel_reviewers.outputs.accuracy_checker.inaccuracies | json }}
-      Outdated: {{ parallel_reviewers.outputs.accuracy_checker.outdated_info | json }}
-      Unverifiable: {{ parallel_reviewers.outputs.accuracy_checker.unverifiable | json }}
-      Recommendations: {{ parallel_reviewers.outputs.accuracy_checker.recommendations }}
-
-      Create a consolidated review with:
-      1. An overall quality score (average of the three scores)
-      2. A prioritized list of changes (critical → minor)
-      3. Any items that need human clarification before proceeding
-      4. A summary suitable for presenting to the document author
-    output:
-      overall_score:
-        type: number
-        description: Average quality score from 1-10
-      critical_changes:
-        type: array
-        description: Must-fix issues (inaccuracies, major clarity problems)
-      important_changes:
-        type: array
-        description: Should-fix issues (readability, structure improvements)
-      minor_changes:
-        type: array
-        description: Nice-to-fix issues (formatting, style tweaks)
-      needs_clarification:
-        type: array
-        description: Items requiring human input or clarification
-      review_summary:
-        type: string
-        description: Human-readable summary of the review
-
-    routes:
-      - to: human_review
-
-  # ── Step 4: Human-in-the-loop review gate ───────────────────────────
-  - name: human_review
-    type: human_gate
-    description: Present review findings to the author for approval or clarification
-    prompt: |
-      ## Document Review Complete
-
-      **Overall Score:** {{ review_synthesizer.output.overall_score }}/10
-
-      ### Review Summary
-      {{ review_synthesizer.output.review_summary }}
-
-      ### Critical Changes (must fix)
-      {% for change in review_synthesizer.output.critical_changes %}
-      - {{ change }}
-      {% endfor %}
-
-      ### Important Changes (should fix)
-      {% for change in review_synthesizer.output.important_changes %}
-      - {{ change }}
-      {% endfor %}
-
-      ### Minor Changes (nice to fix)
-      {% for change in review_synthesizer.output.minor_changes %}
-      - {{ change }}
-      {% endfor %}
-
-      {% if review_synthesizer.output.needs_clarification | length > 0 %}
-      ### ⚠️ Items Needing Clarification
-      {% for item in review_synthesizer.output.needs_clarification %}
-      - {{ item }}
-      {% endfor %}
-      {% endif %}
-    options:
-      - label: Apply All Changes
-        value: apply_all
-        route: document_refiner
-      - label: Apply with Guidance
-        value: apply_with_guidance
-        route: document_refiner
-        prompt_for: feedback
-      - label: Reject & Keep Original
-        value: reject
-        route: $end
-
-  # ── Step 5: Refine the document based on review feedback ────────────
-  - name: document_refiner
-    model: claude-sonnet-4.5
-    description: Applies all approved changes to produce the refined document
-    input:
-      - workflow.input.document
-      - workflow.input.audience
-      - review_synthesizer.output
-      - human_review.output
-      - parallel_reviewers.outputs.accuracy_checker
-    tools: []
-    system_prompt: |
-      You are an expert technical writer and editor. Your task is to apply
-      review feedback to produce a polished, refined version of a document.
-      Preserve the author's voice and intent while improving quality.
-    prompt: |
-      Refine the following document by applying the review feedback.
-
-      **Target Audience:** {{ workflow.input.audience }}
-
-      **Original Document:**
-      {{ workflow.input.document }}
-
-      **Review Score:** {{ review_synthesizer.output.overall_score }}/10
-
-      **Critical Changes to Apply:**
-      {{ review_synthesizer.output.critical_changes | json }}
-
-      **Important Changes to Apply:**
-      {{ review_synthesizer.output.important_changes | json }}
-
-      **Minor Changes to Apply:**
-      {{ review_synthesizer.output.minor_changes | json }}
-
-      **Accuracy Corrections:**
-      Inaccuracies: {{ parallel_reviewers.outputs.accuracy_checker.inaccuracies | json }}
-      Outdated Info: {{ parallel_reviewers.outputs.accuracy_checker.outdated_info | json }}
-
-      {% if human_review.output.feedback is defined and human_review.output.feedback %}
-      **Author Guidance:**
-      {{ human_review.output.feedback }}
-      {% endif %}
-
-      Instructions:
-      1. Apply ALL critical changes (especially factual corrections)
-      2. Apply important changes that improve clarity and readability
-      3. Apply minor changes where they improve polish
-      4. Preserve the author's voice and overall structure
-      5. If the author provided guidance, follow it carefully
-      6. Ensure the document is well-formatted markdown
-      7. Output the complete refined document — not a summary of changes
-    output:
-      refined_document:
-        type: string
-        description: The complete refined markdown document
-      changes_made:
-        type: array
-        description: Summary of changes applied
-      changes_skipped:
-        type: array
-        description: Changes not applied and why
-    routes:
-      - to: final_quality_check
-
-  # ── Step 6: Final quality check ─────────────────────────────────────
-  - name: final_quality_check
-    model: claude-opus-4.6
-    description: Final quality check to ensure the refined document meets standards
-    input:
-      - workflow.input.audience
-      - document_refiner.output.refined_document
-      - review_synthesizer.output.overall_score
-    tools: []
-    prompt: |
-      Perform a final quality check on this refined document.
-
-      **Target Audience:** {{ workflow.input.audience }}
-      **Pre-Refinement Score:** {{ review_synthesizer.output.overall_score }}/10
-
-      **Refined Document:**
-      {{ document_refiner.output.refined_document }}
-
-      Evaluate:
-      1. Were the critical issues from the review addressed?
-      2. Is the document clear, readable, and accurate?
-      3. Is the markdown formatting correct and consistent?
-      4. Is the tone appropriate for the target audience?
-
-      Provide a final quality score and indicate whether the document passes.
-    output:
-      final_score:
-        type: number
-        description: Final quality score from 1-10
-      passed:
-        type: boolean
-        description: Whether the document meets quality standards
-      remaining_issues:
-        type: array
-        description: Any remaining issues (should be minimal)
-    routes:
-      - to: $end
-        when: "{{ output.passed }}"
-      - to: document_refiner
-        when: "context.iteration < 3"
-      - to: $end
-
-parallel:
-  - name: parallel_reviewers
-    description: Run clarity, readability, and accuracy reviews in parallel
-    agents:
-      - clarity_reviewer
-      - readability_reviewer
-      - accuracy_checker
-    failure_mode: continue_on_error
-    routes:
-      - to: review_synthesizer
-
-output:
-  refined_document: "{{ document_refiner.output.refined_document }}"
-  changes_made: "{{ document_refiner.output.changes_made | json }}"
-  original_score: "{{ review_synthesizer.output.overall_score }}"
-  final_score: "{{ final_quality_check.output.final_score }}"
-  review_summary: "{{ review_synthesizer.output.review_summary }}"