{
  "context": "Tests whether the agent correctly processes a multilingual tech conference talk into a rhetoric knowledge vault — covering the 14-dimension analysis format, English-only language policy for non-English quotes, observable-only pattern tagging with the correct scoring formula, and the required structured output files.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "Analysis file location",
      "description": "The per-talk rhetoric analysis file is written inside an `analyses/` subdirectory of the vault root (e.g., vault/analyses/2024-devoxx-jvm-myths.md), not in the vault root itself",
      "max_score": 5
    },
    {
      "name": "All 14 dimensions present",
      "description": "The analysis file contains coverage of all 14 dimensions: Opening Pattern, Narrative Structure, Humor & Wit, Audience Interaction, Transition Techniques, Closing Pattern, Verbal Signatures, Slide-to-Speech Relationship, Persuasion Techniques, Cultural References, Technical Content Delivery, Pacing Clues, Slide Design Patterns, and Areas for Improvement",
      "max_score": 15
    },
    {
      "name": "English-only quote format",
      "description": "At least one Russian phrase from the transcript appears in the analysis with the English translation FIRST, followed by the Russian original in parentheses — e.g., \"That's how it works\" (Вот как это работает) — NOT Russian first or Russian only",
      "max_score": 10
    },
    {
      "name": "Language-tagged verbal signatures",
      "description": "Russian verbal signatures are stored separately from the English signature list and tagged with language code — e.g., [ru] \"ну и\" — NOT merged into the main English verbal signatures section",
      "max_score": 8
    },
    {
      "name": "Structured data fields",
      "description": "The analysis or tracking DB entry includes all required structured_data fields: slide_count, talk_duration_estimate, meme_count, image_only_slide_count, audience_interaction_count, opening_type, closing_type, narrative_arc_type",
      "max_score": 8
    },
    {
      "name": "Pattern score formula",
      "description": "The per-talk pattern_score is calculated as count(patterns) minus count(antipatterns), not a sum or average of the individual pattern values",
      "max_score": 10
    },
    {
      "name": "Observable patterns only",
      "description": "None of the following unobservable patterns appear in the pattern_observations scored section: preparation, carnegie-hall, stakeout, posse, seeding-satisfaction, shoeless, lightsaber, red-yellow-green, laser-weapons, bunker, backchannel",
      "max_score": 8
    },
    {
      "name": "Verbatim examples",
      "description": "The analysis includes a verbatim_examples section with at least three populated categories from: signature_phrases, jokes, transitions, audience_addresses, opening_lines, closing_lines",
      "max_score": 9
    },
    {
      "name": "Tracking DB updated",
      "description": "tracking-database.json exists and contains the processed talk with status set to either 'processed' or 'processed_partial' (not 'pending'), and includes a processed_date value",
      "max_score": 10
    },
    {
      "name": "Areas for improvement present",
      "description": "The analysis file contains a section for dimension 14 (Areas for Improvement / Reflection) that identifies at least two specific weaknesses or risks in the talk's delivery — not just a placeholder, empty section, or single generic statement",
      "max_score": 9
    },
    {
      "name": "Presentation Patterns Scoring section",
      "description": "The analysis file contains a 'Presentation Patterns Scoring' section listing at least one detected pattern or antipattern with confidence level (strong/moderate/weak) and evidence",
      "max_score": 8
    }
  ]
}