Processing Rules

Language Policy — English Only

All analysis output, rhetoric summary updates, tracking DB entries, and profile data MUST be written in English regardless of the talk's delivery language. For non-English talks:

• Verbatim quotes: ALWAYS write English translation FIRST, then the original in parentheses. Never the reverse. Format: "English text" (оригинальный текст). Example: "That's the whole point" (В этом весь смысл) — NOT "В этом весь смысл" (That's the whole point)
• Verbal signatures: store separately tagged with language code (e.g., [ru] "получается что") — do NOT merge into the main English signature list
• Slide text: translate in the analysis, note original language
• Humor/wordplay: note when a joke is language-dependent and untranslatable
• Tag the talk entry with delivery_language in the tracking DB

Pattern Taxonomy Migration

If the pattern taxonomy exists (skills/presentation-creator/references/patterns/_index.md) but any talks with status "processed" or "processed_partial" have no pattern_observations (or pattern_observations.pattern_ids is empty), mark them "needs-reprocessing" with reprocess_reason: "pattern_scoring_added". Report: "N talks need reprocessing for pattern scoring."

Pattern Tagging Rules

Scan observations against the pattern taxonomy index at skills/presentation-creator/references/patterns/_index.md (path relative to plugin root). Skip patterns marked observable: false — these are pre-event logistics and physical stage behaviors that cannot be detected from transcripts or slides. For each observable pattern/antipattern, determine if the talk exhibits it (strong/moderate/weak confidence), record evidence, and compute per-talk pattern score: count(patterns) − count(antipatterns). Return in the pattern_observations field.

Structured Field Extraction

The subagent's job is to return every structured field it identifies (co-presenter, delivery language, slide counts, opening/closing types, etc.) in the structured_data block per the return schema — never to leave them buried only in rhetoric_notes free text. If it's in the analysis, it must be in structured_data.

Persisting those fields is deterministic and script-owned, not a manual per-run mapping — SKILL.md Step 4 runs scripts/persist-results.py for the merge. Authors do not re-derive that logic here.

Adherence Assessment

adherence_assessment measures how consistent a talk is with the speaker's established rhetorical baseline — not whether the talk was good in the abstract. Adherence is consistency with this speaker's own validated style, which is why it can only be computed once a baseline exists.

Gate: produce an assessment only when 10+ scored talks exist — talks with status processed/processed_partial that carry a pattern_score. The assessment anchors to pattern_score vs. the baseline. An unscored talk cannot be assessed. Below that, return "". The subagent reads the baseline from Section 15 of rhetoric-style-summary.md (signature patterns, recurring antipatterns, running average pattern score) — see Rhetoric Summary — Improvement & Adherence Sections below.

Three checks, in order:

1. Pattern adherence — did the talk deploy the speaker's signature patterns and avoid their recurring antipatterns? Underuse counts here too: skipping signature patterns or a narrow range (few distinct patterns) is non-adherence even with zero antipatterns. Anchor to this talk's pattern_score and distinct pattern count versus the baseline — use the talk's mode baseline when Section 15 has a stable one (≥3 talks in that mode), otherwise the global baseline. A lightning talk measured against a keynote baseline produces false "underuse" findings; match like to like.
2. Intent adherence — does the talk honor confirmed intents and design rules, or violate one? A violated confirmed intent is the strongest non-adherence signal.
3. Departure classification — classify each divergence as a deliberate mode-driven choice (different presentation mode, co-presenter, venue) or unintentional backsliding (a recurring antipattern resurfacing). Only backsliding counts against adherence; deliberate departures are noted, not penalized.

Required anchors — the assessment MUST:

• State this talk's pattern_score relative to the running average (e.g., "4 vs. 6.8 average").
• Name any recurring antipattern already tracked in Section 15 that reappeared in this talk.

Bound: 2–4 sentences of prose, not a score — the numeric signal already lives in pattern_observations.pattern_score; the assessment interprets it against the baseline.

Rhetoric Summary — Improvement & Adherence Sections (15–16)

rhetoric-style-summary.md Sections 1–14 mirror the 14 analysis dimensions. Sections 15–16 are cross-talk aggregates, updated in Step 5 each batch.

Section 15 — Improvement & Adherence Baseline

The running baseline that per-talk adherence_assessment measures against. Five required subsections:

1. Recurring improvement themes — issues appearing in 2+ talks. One entry per theme: the issue, the related antipattern ID where one applies (Dimension 14 lists the candidates), severity (hard_limit|warning|info), the count of talks exhibiting it, and the first/last talk filenames where it appeared. Source: aggregate pattern_observations.antipatterns_detected and areas_for_improvement across processed talks.
2. Pattern-score & breadth baseline — running average_pattern_score across scored talks with its trajectory (improving|stable|declining), plus pattern breadth (average distinct patterns per talk) with its trend (widening|stable|narrowing). Track both: a score can decline from antipatterns rising OR from breadth narrowing (using fewer patterns), and these are different coaching messages. Maintain the same figures per presentation mode once a mode has ≥3 scored talks (mirrors the profile's pattern_profile.by_mode); these per-mode figures are what mode-aware adherence compares against. This is the baseline per-talk adherence cites.
3. Signature patterns & strengths — the speaker's high-usage patterns (the adherence reference set). A talk that drops them is a departure to classify; chronic dropping is underuse, not just a one-off. Also surface these as strengths — "lean in / double down" — the positive counterpart to recurring issues, so the baseline isn't purely deficit-oriented. Mirrors the profile's pattern_profile.strengths.
4. Underused patterns (growth) — observable patterns the speaker never or rarely uses that fit their established modes. Framed as range to expand, not a deficiency — the positive-space counterpart to recurring antipatterns. Mirrors the profile's pattern_profile.underused_patterns.
5. Resolved issues — themes that previously recurred but have not appeared in the last 3+ talks. Move an entry here from "recurring themes" once it stops; never delete it — the trajectory is itself signal.

Section 15 is the human-readable source for the profile's pattern_profile and guardrail_sources.recurring_issues (see ../../vault-profile/references/speaker-profile-schema.md); keep the two consistent when both update.

Section 16 — Speaker-Confirmed Intent

Patterns the speaker confirmed as deliberate or accidental during a clarification session (see vault-clarification). Read-only during ingress — populated by clarification, consumed here as the intent-adherence input for Section 15.

Improvement Goal Verification

Each ingress run, after the final Section 15 baseline is current, verifies the speaker's active improvement_goals (set during clarification — record schema in ../../vault-clarification/references/schemas-config.md). This closes the loop: the system stops merely diagnosing and checks whether the issue the speaker chose to work on actually moved.

For each goal with status not in (achieved, retired):

• Compute current_value for the goal's metric from the current Section 15 baseline — and, for pacing and mode-specific goals, from the freshly regenerated speaker profile (this step runs after Step 7, so pacing.adherence and pattern_profile.by_mode are current). Examples by kind: antipattern → the antipattern's frequency over recent talks; underuse → the pattern's recent usage or distinct-pattern breadth; pacing → pacing.adherence.over_budget_rate. Write current_value, last_checked (today), checked_by: "vault-ingress".
• Set status by comparing current_value against baseline_value and target:
  • achieved — current_value meets or beats target.
  • improving — moved toward target versus baseline_value but not there yet.
  • stalled — no meaningful movement from baseline_value.
  • regressed — moved away from target (worse than baseline_value).
• Only count talks processed after set_date toward movement — a goal can't be judged on talks that predate it.
• Never overwrite baseline_value, target, issue, or set_date — those are the fixed yardstick; verification is non-owner and touches status fields only.

Report each goal's status in the run summary. A regressed or stalled goal is the strongest signal to surface — it is the speaker's own priority, not a machine-chosen one.