CtrlK
BlogDocsLog inGet started
Tessl Logo

jbaruch/speaker-toolkit

Six-skill presentation system: ingest talks into a rhetoric vault, run interactive clarification, generate a speaker profile, create presentations that match your documented patterns, produce the deck illustrations + thumbnail visual layer, and publish talk pages to a Jekyll shownotes site. Includes a 111-entry Presentation Patterns taxonomy (81 observable: 62 patterns + 19 antipatterns; 30 unobservable: 21 patterns + 9 antipatterns) for scoring, brainstorming, and go-live preparation.

—

—
Quality

—

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Low

Low-risk findings worth noting

Overview
Quality
Evals
Security
Files

speaker-profile-schema.mdskills/vault-profile/references/

Speaker Profile Schema

The speaker profile (speaker-profile.json) is the machine-readable bridge between the rhetoric-knowledge-vault (analysis) and the presentation-creator skill (generation). It lives in the vault root alongside rhetoric-style-summary.md and slide-design-spec.md.

The narrative summary is the constitution (rich, prose, nuanced). The speaker profile is the specification (structured, extractable, actionable). Both are needed — the profile cannot replace the summary, but the summary alone cannot drive presentation creation at runtime.

When to Generate

  • First generation: After 10+ talks are parsed AND the clarification session has resolved key ambiguities (presentation modes, deliberate vs accidental patterns, confirmed design rules).
  • Updates: After each subsequent vault run (new talks parsed or new PPTX extractions), the profile should be regenerated to incorporate new data.
  • Manual trigger: The user can request "update speaker profile" at any time.

Schema Versioning

Current schema_version: 5. The validator (scripts/validate-profile.py, CURRENT_SCHEMA_VERSION) accepts only the current version.

  • v1 → v2 adds the coaching-outcome fields, all additive: pattern_profile.score_drivers, pattern_breadth, underused_patterns, by_mode, strengths/strengths_note, and pacing.adherence.
  • v2 → v3 makes the Presentation Pattern cohort auditable. It adds the canonical pattern_profile.pattern_baseline snapshot and exact sorted baseline_talk_filenames, and requires every catalog-derived denominator to use that cohort. This is a generation boundary, not an additive coaching field.
  • v3 → v4 binds occurrence history to scoring-v5's exhaustive per-talk outcomes. It separates eligible_talk_count from raw-score-comparable talks_scored, adds exact per-pattern opportunity denominators and coverage, and makes classification availability explicit. Its derived fields remain fail-closed because v4 carries no policy identity.
  • v4 → v5 adds deterministic policy-derived history. It embeds the normalized policy plus its canonical semantic digest, adds exhaustive positive and antipattern classification rows, makes availability independent per derived domain, and retains the complete trend audit used by projections.
  • Generation: vault-profile writes only v5. Stored v1/v2/v3/v4 profiles are non-current and scripts/validate-profile.py rejects them as generation output. A compatible reader may expose a validated v4 profile's occurrence rows only; every v4 mastery, novelty, severity, combination, underuse, and trend field remains unavailable because v4 has no policy stamp.
  • Migration: vault-profile regenerates the profile wholesale each run. An older file is replaced by a v5 file on the next run — no in-place migration step and no talk reparse. Existing current-generation tracking rows are read without mutation. The only value carried across regenerations is infrastructure.template_layouts[].use_for (merged by the (master_index, name) pair, version-independent).
  • Generation reset: when either the catalog fingerprint or pattern-scoring schema changes, catalog-derived diffs across the boundary are a reset. Do not describe score, usage, mastery, or trend changes across those identities as speaker regressions. Within one generation, a changed classification-policy semantic digest resets only policy-derived comparisons; raw occurrence rows remain comparable.

Schema

{
  "schema_version": 5,
  "generated_date": "2026-02-22",
  "talks_analyzed": 24,

  "speaker": {
    "name": "",
    "handle": "",
    "website": "",
    "bio_short": "one-sentence bio used on slides",
    "bio_context": "career trajectory or credentials chain shown on bio slides"
  },

  "infrastructure": {
    "template_pptx_path": "/path/to/template.pptx",
    "template_name": "human-readable name for the template",
    "template_layouts": [
      {
        "index": 0,
        "master_index": 0,
        "name": "TITLE",
        "placeholders": [{"idx": 0, "type": "CENTER_TITLE"}],
        "use_for": "opening title slide, section dividers"
      }
    ],
    // template_layouts: structural fields (index, master_index, name,
    // placeholders) come from skills/vault-ingress/scripts/pptx-extraction.py.
    // The use_for field is speaker-curated and persists across
    // regenerations; the vault-profile aggregator merges fresh structural
    // data with prior use_for values by keying on the (master_index, name)
    // pair — name alone is unsafe because PowerPoint permits identical
    // layout names under different masters.
    "presentation_file_convention": "{presentations_dir}/{conference}/{year}/{talk-slug}/",
    "font_pair": {
      "title": {"name": "", "source": "google_fonts|system|custom"},
      "body": {"name": "", "source": "google_fonts|system|custom"},
      "code": {"name": "", "source": "google_fonts|system|custom", "optional": true}
    },
    "slide_dimensions": {"width_inches": 10.0, "height_inches": 5.63, "aspect": "16:9"}
  },

  "presentation_modes": [
    {
      "id": "a",
      "name": "human-readable mode name",
      "description": "1-2 sentence description of this mode",
      "when_to_use": "what spec signals suggest this mode",
      "slide_density_per_min": 1.4,
      "meme_density_per_slide": 0.24,
      "humor_register": "none|light|moderate|heavy",
      "audience_interaction": true,
      "anti_sell_applicable": true,
      "commercial_intent": "none|subtle|direct",
      "profanity_default": "zero|moderate|heavy",
      "closing_default": "name of default closing pattern for this mode",
      "typical_engine": "pptx|presenterm — optional; the engine this mode usually renders in"
    }
  ],

  "presentation_engines": [
    {
      "id": "pptx",
      "name": "human-readable engine name",
      "renderer": "pptx|presenterm",
      "when_to_use": "what spec signals suggest this engine",
      "default_theme": "theme/template name for this engine, or null",
      "usage_count": 18,
      "out_of": 24
    }
  ],

  "design_rules": {
    "background_color_strategy": "random_non_repeating|theme_sequence|mode_dependent",
    "background_color_pool": "description of which colors to pick from",
    "background_adjacent_repeat": false,
    "white_black_reserved_for": "full-bleed image/meme slides only",
    "slide_numbers": "never|always|optional",
    "footer": {
      "always_present": true,
      "pattern": "the footer template string with placeholders",
      "elements": ["@handle", "#conference", "#topic", "website"],
      "co_presented_extra": "co-presenter handle position and rule",
      "font": "",
      "font_size_pt": 16,
      "position": {"left": 0.01, "top": 5.22, "width": 10.0, "height": 0.37},
      "color_adapts_to_background": true,
      "outline_for_legibility": true
    },
    "memes_always_full_bleed": true,
    "default_bullet_symbol": "multiplication_sign|dash|circle|custom",
    "contextual_bullet_symbols": true,
    "corporate_watermark": "never|always|conditional",
    "section_dividers": "text_cue|numbered_slide|color_change|none"
  },

  "rhetoric_defaults": {
    "default_duration_minutes": 45,
    "modular_design": true,
    "default_opening": "description of the default opening pattern",
    "delayed_self_intro": {"enabled": true, "brief_bio_slide": 3, "full_reintro": "mid-talk"},
    "profanity_calibration": "per_audience|fixed|none",
    "on_slide_profanity": "never_default|needs_explicit_approval",
    "anti_sell_pattern": true,
    "three_part_close": true,
    "shownotes_slide_position": "early (slide 4-5)"
  },

  "confirmed_intents": [
    {
      "schema_version": 1,
      "pattern": "name of the pattern",
      "intent": "deliberate",
      "rule": "what the presentation-creator should do about it",
      "note": "additional context from the speaker"
    }
  ],

  "pacing": {
    "wpm_range": {"min": 127, "max": 162, "comfortable": 135},
    "slides_per_minute": {"comfortable": 1.4, "max": 1.5},
    "meme_section_pace": "30-40 sec/slide",
    "data_section_pace": "60-90 sec/slide",
    "demo_pace": "minimal slides, live tool is the content",
    "adherence": {
      "cohort": "current_instrumentation_talks",
      "talks_over_budget": 5,
      "talks_scored": 24,
      "over_budget_rate": 0.21,
      "trend": "improving|stable|worsening",
      "worst_offenders": [
        {"filename": "2024-04-10-talk-slug.md", "slides_per_minute": 2.1, "budget_slides_per_minute": 1.5, "over_by": "40%"}
      ],
      "note": "Quantitative time/slide pacing, computed by scripts/compute-pacing-adherence.py from the separately named current_instrumentation_talks cohort. It uses each talk's structured_data.slide_count and structured_data.talk_duration_estimate vs guardrail_sources.slide_budgets. This is not the Presentation Pattern cohort and its talks_scored may differ from pattern_profile.talks_scored. Duration parsing and budget-band selection live in that script's docstring. Distinct from the qualitative 'rushing' read in vault Dimension 14 (transcript-evident time panic) — this is the corpus-level count. The duration estimate is transcript-derived and approximate; treat marginal overages as soft signals, not hard failures."
    }
  },

  "guardrail_sources": {
    "slide_budgets": [
      {"duration_min": 20, "max_slides": 30, "slides_per_min": 1.5},
      {"duration_min": 30, "max_slides": 45, "slides_per_min": 1.5},
      {"duration_min": 45, "max_slides": 70, "slides_per_min": 1.5},
      {"duration_min": 60, "max_slides": 90, "slides_per_min": 1.5},
      {"duration_min": 75, "max_slides": 110, "slides_per_min": 1.5},
      {"duration_min": 90, "max_slides": 130, "slides_per_min": 1.4}
    ],
    "act1_ratio_limits": [
      {"duration_range": "20-30 min", "max_percent": 40},
      {"duration_range": "45 min", "max_percent": 45},
      {"duration_range": "60+ min", "max_percent": 50},
      {"duration_range": "75+ min", "max_percent": 65}
    ],
    "recurring_issues": [
      {
        "id": "short_identifier",
        "source_lane": "non_pattern",
        "description": "what tends to go wrong",
        "guardrail": "specific check or rule to prevent it",
        "severity": "hard_limit|warning|info"
      }
    ]
  },

  "instrument_catalog": {
    "opening_patterns": [
      {
        "code": "a",
        "name": "",
        "best_for": "when to use this pattern",
        "description": "1-2 sentences on how it works",
        "frequency": 0
      }
    ],
    "narrative_structures": [
      {
        "name": "",
        "best_for": "which modes/contexts",
        "acts": "brief act breakdown",
        "time_allocation": "percentage split"
      }
    ],
    "humor_techniques": [
      {
        "name": "",
        "register": "safe|moderate|heavy|venue_specific",
        "description": ""
      }
    ],
    "audience_interactions": [
      {"name": "", "best_for": "", "description": ""}
    ],
    "transition_techniques": [
      {"name": "", "description": ""}
    ],
    "closing_patterns": [
      {"name": "", "best_for": "", "structure": ""}
    ],
    "persuasion_techniques": [
      {"name": "", "category": "exposing_problems|building_credibility|selling|creating_frameworks", "description": ""}
    ],
    "verbal_signatures": [
      {"phrase": "", "usage": "when/how it's used", "frequency": "high|medium|low|rare"}
    ],
    "pop_culture_notes": "general guidance on how this speaker uses pop-culture references"
  },

  "visual_style_history": {
    "default_illustration_style": "the speaker's most common illustration aesthetic, or null if no pattern",
    "default_image_source": "most common image source type: ai_generated, meme, screenshot, stock_photo, custom_artwork, mixed",
    "style_departures": [
      {
        "style": "name of the departure style (e.g., retro_tech_manual)",
        "trigger": "what caused the departure: mode, co-presenter, topic, venue",
        "talks": ["list of talk filenames that used this style"],
        "description": "human-readable description of the style, specific enough to inform a prompt"
      }
    ],
    "mode_visual_profiles": [
      {
        "mode_id": "a",
        "typical_style": "illustration aesthetic typically used in this mode",
        "image_source_mix": "what image sources dominate in this mode",
        "notes": "any mode-specific visual tendencies"
      }
    ],
    "evolution_notes": "narrative of how the speaker's visual style has changed over time",
    "visual_continuity_patterns": ["recurring devices across talks: numbering schemes, mascots, progressive elements"],
    "confirmed_visual_intents": [
      {
        "pattern": "name of the visual pattern",
        "intent": "deliberate",
        "rule": "what the presentation-creator should do about it",
        "note": "additional context from the speaker"
      }
    ]
  },

  "pattern_profile": {
    "pattern_baseline": {
      "schema_version": 2,
      "as_of": "2026-02-22T12:00:00+00:00",
      "scope": "global",
      "active_batch_excluded": false,
      "excluded_filenames": [],
      "eligible_statuses": ["processed", "processed_partial"],
      "pattern_scoring_generation_status": "current",
      "pattern_scoring_generation_reasons": [],
      "pattern_catalog_fingerprint": "0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef",
      "pattern_scoring_schema_version": 5,
      "scored_talk_count": 24,
      "pattern_score_sum": 163,
      "average_pattern_score": 6.79,
      "eligible_talk_count": 24,
      "opportunity_coverage_identity": "abcdef0123456789abcdef0123456789abcdef0123456789abcdef0123456789",
      "raw_score_comparison_status": "available",
      "raw_score_comparison_reason": null
    },
    "baseline_talk_filenames": [
      "example-01.md",
      "example-02.md",
      "example-03.md",
      "example-04.md",
      "example-05.md",
      "example-06.md",
      "example-07.md",
      "example-08.md",
      "example-09.md",
      "example-10.md",
      "example-11.md",
      "example-12.md",
      "example-13.md",
      "example-14.md",
      "example-15.md",
      "example-16.md",
      "example-17.md",
      "example-18.md",
      "example-19.md",
      "example-20.md",
      "example-21.md",
      "example-22.md",
      "example-23.md",
      "example-24.md"
    ],
    "eligible_talk_count": 24,
    "talks_scored": 24,
    "average_pattern_score": 6.79,
    "score_trend": "improving",
    "pattern_breadth": {
      "avg_distinct_patterns_per_talk": 6.3,
      "trend": "widening",
      "note": "Breadth is the mean count of detected positive catalog patterns per current-generation talk."
    },
    "underused_patterns": [],
    "score_drivers": {
      "direction": "improving",
      "antipattern_drivers": ["going-meta"],
      "pattern_drivers": [],
      "note": "Classifier-produced drivers for the available trend domain."
    },
    "by_mode": [],
    "strengths": ["narrative-arc"],
    "strengths_note": "Deterministic projection of regular and signature positive-pattern classifications.",
    "note": "Only observable patterns are included. Patterns marked observable: false in the taxonomy (pre-event logistics, hidden authoring/provenance processes, physical stage behaviors, post-event follow-up, and external systems the current artifacts cannot prove) are excluded from scoring and surfaced as a go-live checklist in creator Phase 6 instead.",
    // Opportunity arrays are abbreviated here for readability. Generated
    // profiles contain one sorted row for every observable catalog entry of
    // the matching polarity.
    "pattern_usage": [
      {
        "pattern_id": "narrative-arc",
        "detected_count": 22,
        "evaluable_count": 22,
        "unevaluable_count": 2,
        "not_applicable_count": 0,
        "eligible_cohort_count": 24,
        "coverage": 0.9166666666666666,
        "times_used": 22,
        "out_of": 22,
        "usage_rate": 1.0
      }
    ],
    "antipattern_frequency": [
      {
        "pattern_id": "going-meta",
        "detected_count": 8,
        "evaluable_count": 24,
        "unevaluable_count": 0,
        "not_applicable_count": 0,
        "eligible_cohort_count": 24,
        "coverage": 1.0,
        "times_detected": 8,
        "out_of": 24,
        "frequency_rate": 0.3333333333333333
      }
    ],
    "never_used_patterns": [],
    "signature_combinations": [],
    "mastery_levels": {
      "signature": ["narrative-arc"],
      "regular": [],
      "occasional": [],
      "rare": [],
      "never_tried": []
    },
    "classification_schema_version": 1,
    "classification_policy": {
      "schema_version": 1,
      "policy_id": "speaker-toolkit-default",
      "policy_version": 1,
      "source": "bundled_default",
      "semantic_sha256": "<classifier-produced 64-character lowercase hex digest>",
      "semantic_policy": {
        // Complete classifier-produced object omitted from this illustrative profile.
      }
    },
    "classification_availability": {
      "schema_version": 2,
      "mastery_and_novelty": {"status": "available", "reason_codes": []},
      "antipattern_recurrence": {"status": "available", "reason_codes": []},
      "underuse": {"status": "available", "reason_codes": []},
      "signature_combinations": {"status": "available", "reason_codes": []},
      "trends": {"status": "available", "reason_codes": []},
      "modes": {"status": "unavailable", "reason_codes": ["talk_mode_assignments_unavailable"]}
    },
    // Both arrays below are exhaustive and catalog-sorted in generated profiles;
    // one representative row from each polarity is shown here.
    "pattern_classifications": [
      {
        "pattern_id": "narrative-arc",
        "classification": "signature",
        "observation_status": "observed",
        "absence_conclusion_capable": false,
        "evidence": {
          "applicable_count": 24,
          "evaluable_count": 22,
          "detected_count": 22,
          "unevaluable_count": 2,
          "applicable_coverage": 0.9166666666666666,
          "lower": 0.9166666666666666,
          "upper": 1.0
        },
        "reason_codes": ["meets_signature_thresholds"]
      }
    ],
    "antipattern_classifications": [
      {
        "pattern_id": "going-meta",
        "classification": "moderate_frequency",
        "observation_status": "observed",
        "absence_conclusion_capable": true,
        "evidence": {
          "applicable_count": 24,
          "evaluable_count": 24,
          "detected_count": 8,
          "unevaluable_count": 0,
          "applicable_coverage": 1.0,
          "lower": 0.3333333333333333,
          "upper": 0.3333333333333333
        },
        "reason_codes": ["meets_moderate_frequency_thresholds"]
      }
    ],
    // Angle-bracket values in trend_analysis document output shape only.
    // Copy the classifier-produced values unchanged.
    "trend_analysis": {
      "status": "available",
      "reason_codes": [],
      "sample": {
        "required_talk_count": "<classifier-produced integer>",
        "valid_date_talk_count": "<integer>",
        "invalid_date_filenames": [],
        "selected_filenames": ["<classifier-selected filename>", "..."],
        "opportunity_coverage_identity": "abcdef0123456789abcdef0123456789abcdef0123456789abcdef0123456789"
      },
      "score": {"status": "improving", "prior_average": 6.0, "recent_average": 7.0, "delta": 1.0},
      "breadth": {"status": "widening", "prior_average": 5.0, "recent_average": 6.0, "delta": 1.0},
      // Movement arrays are exhaustive when trends are available and empty when
      // the trend domain is unavailable. One representative row is shown per lane.
      "pattern_movements": [
        {
          "pattern_id": "narrative-arc",
          "movement": "indeterminate",
          "prior_evidence": {"applicable_count": "<integer>", "evaluable_count": "<integer>", "detected_count": "<integer>", "unevaluable_count": "<integer>", "applicable_coverage": "<number or null>", "lower": "<number or null>", "upper": "<number or null>"},
          "recent_evidence": {"applicable_count": "<integer>", "evaluable_count": "<integer>", "detected_count": "<integer>", "unevaluable_count": "<integer>", "applicable_coverage": "<number or null>", "lower": "<number or null>", "upper": "<number or null>"},
          "reason_codes": ["uncertainty_spans_movement_threshold"]
        }
      ],
      "antipattern_movements": [
        {
          "pattern_id": "going-meta",
          "movement": "decreasing",
          "prior_evidence": {"applicable_count": "<integer>", "evaluable_count": "<integer>", "detected_count": "<integer>", "unevaluable_count": "<integer>", "applicable_coverage": "<number or null>", "lower": "<number or null>", "upper": "<number or null>"},
          "recent_evidence": {"applicable_count": "<integer>", "evaluable_count": "<integer>", "detected_count": "<integer>", "unevaluable_count": "<integer>", "applicable_coverage": "<number or null>", "lower": "<number or null>", "upper": "<number or null>"},
          "reason_codes": ["conservative_interval_decrease"]
        }
      ]
    }
  },

  "badges": [
    {
      "id": "short_identifier",
      "source_lane": "non_pattern",
      "name": "Badge display name",
      "description": "What this badge represents — fun, self-deprecating, grounded in vault data",
      "evidence": "specific data point(s) from the vault that earned this badge"
    }
  ],

  "publishing_process": {
    "shownotes": {
      "enabled": true,
      "source": {
        "type": "local_jekyll|local_hugo|local_eleventy|local_astro|remote_url|none",
        "path_or_url": "/path/to/shownotes-site-root (or a remote https URL for remote_url)",
        "talks_subdir": "_talks"
      },
      "url": {
        "base": "https://speaking.example.com",
        "template": "/{slug}/"
      },
      "thumbnail_path_template": "assets/images/thumbnails/{slug}-thumbnail.png",
      "slug_convention": {
        "template": "{venue-compact}{yy}-{short-id}",
        "examples": ["jfokus26-monkey", "devnexus26-robocoders"]
      },
      "ssg_template_pointer": "{source.path_or_url}/_layouts/default.html",
      "publishing_method": "description of how shownotes are published (git push, CMS, manual)",
      "shownotes_template": "path to the SSG template file for new talk pages, or null"
    },
    "export_format": "pdf|pptx_only|both",
    "export_method": "description of how to export (e.g., PowerPoint AppleScript, LibreOffice CLI, manual)",
    "export_script": "optional: literal script/command to run for export, or null",
    "qr_code": {
      "enabled": true,
      "target": "shownotes_url|custom_url",
      "custom_url": "a full https URL (only when target=custom_url)",
      "insert_into_deck": true,
      "slide_position": "shownotes_slide|closing|both",
      "shortener": "bitly|rebrandly|none",
      "bitly_domain": "jbaru.ch | null",
      "rebrandly_domain": "jbaru.ch | null",
      "bg_color_match": true
    },
    "thumbnail": {
      "enabled": true,
      "speaker_photo_path": "/path/to/headshot.jpg",
      "aesthetic_preference": "photo|comic_book",
      "style_preference": "slide_dominant|split_panel|overlay",
      "title_position": "top|bottom|overlay",
      "brand_colors": ["#hex1", "#hex2"],
      "notes": "speaker-specific thumbnail preferences"
    },
    "video_publishing": {
      "enabled": true,
      "embed_method": "youtube_embed|link_only|both",
      "shownotes_video_section": "where/how video goes in shownotes",
      "video_description_template": "template with {conference} {year} {shownotes_url} placeholders"
    },
    "additional_steps": [
      {
        "name": "step name",
        "description": "what to do",
        "automated": true,
        "script": "optional command/script, or null"
      }
    ],
    "notes": "any speaker-specific publishing quirks or preferences"
  }
}

Confirmed-intent intent values are non-empty speaker-owned classification labels, not a closed three-value enum. deliberate, accidental, and context_dependent are recommended defaults, while established labels such as accepted_tradeoff, fact, or deliberate_signature remain authoritative.

The confirmed_intents[].schema_version shown in the profile is the independent public profile-domain intent schema, not the tracking-database record version. load-vault.py projects only pattern, intent, rule, and note from stored intent records; it strips database schema and capture provenance before profile generation. Storage-only fields such as confirmed_date, source-talk pointers, and retrofit_targets intentionally do not enter the semantic profile. The profile producer then emits the profile-domain schema-v1 intent object shown above.

Pattern Cohort Provenance

pattern_profile.pattern_baseline is copied unchanged from the loader's pattern_baseline payload. It uses the vault-ingress adherence-baseline schema and must describe the active catalog fingerprint and scoring schema with active_batch_excluded: false and excluded_filenames: []. For scoring schema v5, the loader also requires every selected talk's persisted source-location ledger and citations to remain fresh against the live vault and configured source roots, plus one canonical exhaustive pattern_outcomes matrix and matching opportunity_coverage_identity. Generation identity alone cannot authorize the cohort. baseline_talk_filenames is the sorted, unique list of exact filenames from baseline_talks; its length and pattern_profile.eligible_talk_count equal pattern_baseline.eligible_talk_count. That eligible count is also copied into every opportunity row. A row's out_of is its own evaluable count, not the global cohort count. pattern_usage and antipattern_frequency contain one sorted row for every observable catalog entry of the matching polarity, including when no talk supplied an evaluable opportunity. Their exact fields and arithmetic are owned by scripts/pattern_opportunities.py.

Raw score fields are a separate lane. talks_scored and average_pattern_score exactly mirror the validated baseline's raw-score-comparison count and average. When eligible talks have mixed opportunity identities, or when their shared identity has no evaluable (detected or undetected) opportunity, the baseline explicitly reports raw_score_comparison_status: "unavailable", zero scored talks, and a null average. The respective stable reasons are mixed_opportunity_coverage and no_evaluable_pattern_opportunities; the per-pattern occurrence rows remain available.

All pattern_profile fields in the schema are required in v5. If the vault has no pattern-classification-policy.json, the loader automatically applies the bundled speaker-toolkit-default@1 policy; users are not asked to invent thresholds. A present override is strict and fail-closed: an unreadable file, duplicate key, non-finite number, unknown/missing field, unsupported version, oversized file, or invalid threshold aborts profile generation rather than silently selecting the default.

classification_policy is self-contained. Copy semantic_policy and semantic_sha256 from classifier output unchanged. Treat the digest as an opaque comparison identity. policy_id, policy_version, and source identify the applied policy for humans.

pattern_classifications and antipattern_classifications contain one sorted row for every observable catalog entry of the matching polarity. Each row preserves the exact evaluable denominator and exposes classifier-produced evidence counts, coverage, and conservative bounds. Positive and antipattern rows use the label vocabularies shown in the schema example; row-level unclassified results remain part of the exhaustive output.

Classifier-Owned Policy Semantics

Thresholds, formulas, tier predicates, combination selection, trend windows, and movement decisions are not restated in this reference. See references/pattern-classification-policy-v1.json for the complete bundled semantic policy and scripts/classify-pattern-profile.py — validate_policy() and classify_pattern_profile() — for its executable contract. Each generated profile embeds the normalized policy that was actually applied, plus its semantic digest. Consumers copy the classifier's rows and projections unchanged.

mastery_levels, strengths, underused_patterns, never_used_patterns, signature combinations, and trend analysis are deterministic classifier projections. Consumers must not reconstruct them from raw rates or from the prose summary.

never_tried and not_yet_observed are deliberately not synonyms. never_tried is a policy-backed absence classification. not_yet_observed says that this corpus has no positive detection while absence remains unknown. A conclusive absence that does not reach a named tier remains unclassified with observation status confirmed_absent. Only classifier-emitted never_tried IDs enter never_used_patterns; no other zero state may be presented as proof that the speaker has never tried the technique. The antipattern equivalent is confirmed_none; every other zero-detection antipattern remains non-recurring unless the classifier says otherwise.

classification_availability is schema v2 and independent per domain. Mastery/novelty, antipattern recurrence, underuse, combinations, trends, and modes each carry their own {status, reason_codes}. The classifier output is the only authority for those values. A consumer must gate only the requested domain and retain row-level unclassified results. One unavailable domain never erases another available one. trend_analysis retains the complete selected sample, metric values, exhaustive pattern movements, and reasons behind those projections.

A validated schema-v4 profile remains compatible only as occurrence history: pattern_baseline, filenames, raw score availability, pattern_usage, and antipattern_frequency may be read under their v4 provenance contract. Because v4 has no policy stamp, readers must ignore its mastery, novelty, recurrence, strength, underuse, combination, mode, and trend projections. Regenerate to v5 from the tracking database to obtain those fields; this re-analysis does not reparse talks or modify raw talk/opportunity rows.

talks_analyzed may count every processed talk. Pacing uses the separately named current_instrumentation_talks cohort and never borrows the pattern denominator.

pattern_profile is the only v5 storage lane for catalog history. rhetoric_defaults must not duplicate mastery, signatures, usage, scores, or other pattern-history fields. guardrail_sources.recurring_issues[] and badges[] are non-pattern-only lanes: every entry carries source_lane: "non_pattern" and must not carry pattern IDs, scores, mastery, catalog identity, occurrence fields, or a pattern denominator. Consumers derive catalog warnings and reinforcement directly from the validated pattern_profile; the writer never materializes duplicate catalog-derived recurring issues or badges.

Writers and readers use the shared profile provenance/classification validator for the same strict decision. Current v5 catalog availability requires source-exact cohort and opportunity rows; each classification domain then follows its own availability object. The owner additionally runs "{python_path}" "{speaker_toolkit_root}/skills/vault-profile/scripts/validate-profile.py" --vault-root <path>; it recomputes the cohort from the live tracking database and rejects structurally valid but source-fabricated occurrence, policy, classification, availability, or projection rows. A reader may continue using unrelated non-pattern fields from an older profile and the explicitly supported v4 occurrence-only lane above, but it must treat all other legacy pattern history as unavailable rather than migrate or infer provenance.

When the current pattern cohort is empty, preserve the canonical zero-count baseline, an empty baseline_talk_filenames, eligible_talk_count: 0, talks_scored: 0, and average_pattern_score: null. Keep the exhaustive positive and negative catalog rows with zero counts and null rates/coverage, plus exhaustive classification rows with their classifier-owned unavailable/unclassified evidence. Set score_trend, pattern_breadth.trend, and score_drivers.direction to "unavailable"; set the breadth average to null and derived list/mastery projections to []. Never fall back to an older profile, the rhetoric summary, excluded scoring generations, or the instrumentation cohort.

How the Presentation Creator Uses Each Section

The presentation-creator skill reads the profile at runtime. Its reference files define the PROCESS; the profile provides the DATA. No generation step is needed — the creator automatically picks up changes when the profile is regenerated.

Profile sectionCreator phase that reads itWhat it drives
speaker + infrastructurePhase 0 (load)Vault path, template reference, file conventions
presentation_modesPhase 2 (architecture)Mode selection menu and recommendations
presentation_enginesPhase 2 (Decision #2 — engine & theme sourcing), Phase 5 (slide generation)Engine sourcing menu; Phase 5 reads the chosen renderer. Optional/additive — absent on older profiles, which fall back to a flat pptx/presenterm menu
design_rulesPhase 5 (slide generation)Background colors, footer specs, shape vocabulary
rhetoric_defaultsPhase 1-3 (spec, architecture, content)Voice calibration, opening/closing defaults
confirmed_intentsPhase 2-4 (architecture, guardrails)Hard rules that override pattern inference
pacingPhase 3-4 (content, guardrails)Slide budget tables, WPM targets
guardrail_sourcesPhase 4 (guardrails)All guardrail checks with thresholds
instrument_catalogPhase 2 (architecture)Complete instrument menu by dimension
visual_style_historyPhase 2 (architecture — illustration strategy)Default aesthetic, mode-specific departures, style proposals
pattern_profilePhase 2 (architecture), Phase 4 (guardrails)Auditable occurrence and exhaustive classification rows; recommendations and warnings gated by the relevant availability domain and row evidence
badgesInformationalFun speaker achievements mined from vault data
infrastructure.template_layoutsPhase 5 (slide generation)Layout map and selection logic
infrastructure.font_pairPhase 5 (slide generation)Font usage rules
publishing_processPhase 6 (publishing), Phase 7 (post-event)Export, shownotes, QR code, distribution steps, thumbnail prefs, video publishing

skills

vault-profile

references

pattern-classification-policy-v1.json

profile-construction-rules.md

schemas-config.md

speaker-profile-schema.md

SKILL.md

README.md

tile.json