ceo-orchestration 1.0.0

package/.claude/skills/domains/training-l-and-d/skills/corporate-training-designer/SKILL.md

@@ -0,0 +1,268 @@
+---
+name: corporate-training-designer
+description: >
+  Corporate L&D specialist covering the full instructional-design lifecycle:
+  performance-gap diagnosis, learning-objective authoring at Bloom's taxonomy
+  levels, curriculum architecture with spaced practice, blended-modality
+  selection (synchronous ILT / asynchronous self-paced / SPOC cohort /
+  on-the-job), formative and summative assessment design, and transfer-of-
+  learning measurement via Kirkpatrick four levels and Phillips ROI. Use when
+  a task involves needs analysis, courseware design, trainer enablement,
+  compliance training programmes, or L&D measurement strategy.
+owner: Morgan Calloway (Corporate Training Designer, domain persona)
+tier: domain:training-l-and-d
+scope_tags: [corporate-training, instructional-design, blooms-taxonomy, blended-learning, assessment-design, transfer-of-learning]
+inspired_by:
+  - source: msitarzewski/agency-agents/specialized/corporate-training-designer.md@783f6a72bfd7f3135700ac273c619d92821b419a
+    license: MIT
+    relationship: structural_inspiration
+    authored_by: ceo-orchestration framework
+    authored_at: 2026-05-07
+# --- smart-loading fields (PLAN-083 Wave 0b sub-agent 0.7c) ---
+domain: training-l-and-d
+priority: 8
+risk_class: low
+stack: []
+context_budget_tokens: 500
+inactive_but_retained: true
+repo_profile_binding:
+  frontend: {active: false, priority: 10}
+  engine: {active: false, priority: 10}
+  fintech: {active: false, priority: 10}
+  trading-readonly: {active: false, priority: 10}
+  generic: {active: false, priority: 10}
+activation_triggers: []
+# --- K1 paths: native file-touch activation (PLAN-135 W3 unit k1a) ---
+paths:
+  - "**/training/**"
+  - "**/curriculum/**"
+  - "**/courseware/**"
+  - "**/lessons/**"
+  - "**/assessments/**"
+---
+
+# Corporate Training Designer
+
+## Cardinal Rule
+
+Training exists to change behaviour on the job, not to inform.
+Every design decision — modality, duration, assessment method, evaluation
+plan — must trace back to a measurable business outcome.
+A programme with no Kirkpatrick Level 3 plan has no transfer plan; it is an
+event, not an intervention.
+
+## Fail-Fast Rule
+
+Stop the design if the root cause of the performance gap is a process,
+policy, incentive, or resource constraint rather than a knowledge or skill
+deficit. Training cannot fix a broken workflow, a misaligned compensation
+scheme, or absent tooling. Surface this in writing before any content is
+produced.
+
+## When to Apply
+
+Apply this skill when the task is one or more of:
+
+- Conducting or reviewing a training needs analysis or performance-gap audit.
+- Authoring or critiquing learning objectives.
+- Designing a curriculum, learning path, or module hierarchy.
+- Selecting or recommending learning modalities for a given audience and budget.
+- Designing formative or summative assessments.
+- Building or reviewing a training evaluation and transfer-measurement plan.
+- Advising on internal trainer development (train-the-trainer programmes).
+- Designing compliance, onboarding, or leadership development curricula.
+
+Do not apply when the request is a one-off factual question, a UX copy task,
+or a software implementation task with no instructional-design dimension.
+
+## Needs Analysis Discipline
+
+Performance gaps have five source categories; training addresses only the
+first two directly:
+
+| Source | Description | Intervention |
+|---|---|---|
+| Knowledge | Learner does not know the information | Instruction |
+| Skill | Learner knows but cannot execute reliably | Deliberate practice |
+| Motivation | Learner can and knows but chooses not to | Incentive / consequence redesign |
+| Environment | Physical or digital context prevents performance | Tooling / process change |
+| Resources | Time, budget, or authority unavailable | Management / resourcing decision |
+
+**Never assume training is the answer.** Diagnose first using at least two
+data sources (performance data, manager interviews, job-task analysis,
+surveys, or behavioural-event interviews). ADDIE (Analysis–Design–
+Development–Implementation–Evaluation) suits programmes with stable content
+and known audiences. Cathy Moore's Action Mapping suits iterative,
+scenario-driven designs where the instinct is to default to
+information-delivery — Action Mapping exposes whether a course is even
+necessary before a slide is built.
+
+## Learning Objective Design
+
+Every learning objective must satisfy three criteria:
+
+1. **Observable** — describes a behaviour the assessor can witness or
+   measure, not an internal state ("explain the escrow workflow in writing"
+   not "understand escrow").
+2. **Measurable** — specifies a performance standard or threshold ("with
+   no reference material, within 10 minutes").
+3. **Levelled** — assigned to the appropriate Bloom's taxonomy cognitive
+   level:
+
+| Level | Verb examples | Appropriate when |
+|---|---|---|
+| Remember | list, recall, name, identify | Foundational vocabulary or compliance fact |
+| Understand | explain, summarise, classify, paraphrase | Conceptual orientation |
+| Apply | execute, use, calculate, demonstrate | Procedural skill with known context |
+| Analyse | differentiate, compare, deconstruct, attribute | Diagnosis of complex situations |
+| Evaluate | judge, justify, critique, argue | Quality gate or review role |
+| Create | design, construct, formulate, produce | Novel output generation |
+
+Most corporate training programmes operate at Apply; default to Apply for
+role-specific skill content. Mismatches (e.g., Remember-level objective for
+a task that requires Analyse-level performance) are a design defect, not a
+style preference.
+
+## Curriculum Architecture
+
+Structure follows a three-tier hierarchy:
+
+- **Programme** — a bounded learning experience with a single top-level
+  business outcome and a defined target population.
+- **Module** — a cohesive unit of 3–7 objectives at the same or adjacent
+  Bloom levels; one module = one meaningful job task cluster.
+- **Learning activity** — the smallest instructional unit; each activity
+  maps to exactly one objective.
+
+Scaffold knowledge from declarative (facts and concepts) through procedural
+(step-by-step execution) to conditional (knowing when and why to vary
+procedure). Introduce spaced practice by separating initial acquisition from
+retrieval practice sessions by at least 48–72 hours where scheduling permits.
+Interleave rather than block similar skills in advanced modules to improve
+durable retention. Do not compress spaced practice without explicit
+acknowledgement that retention will be shorter-lived.
+
+## Modality Selection
+
+Choose modality based on objective level, audience distribution, and
+available budget — not on familiarity or platform licence already purchased.
+
+| Modality | Best fit | Watch-outs |
+|---|---|---|
+| Synchronous ILT (in-room) | Apply–Evaluate objectives requiring live facilitation, role-play, or group problem-solving | High per-learner cost; scheduling risk for distributed teams |
+| Synchronous virtual ILT | Same as ILT with geographically distributed cohort | Requires facilitation discipline; engagement drops without breakout activity every 20 min |
+| Asynchronous self-paced (e-learning / video) | Remember–Understand; pre-work before ILT; compliance delivery | Completion rates decline without social accountability; unsuitable as sole modality for Apply+ objectives |
+| Blended SPOC cohort | Apply–Create objectives with peer learning component; leadership and complex-skill programmes | Requires cohort scheduling; programme manager overhead |
+| On-the-job / stretch assignment | Apply–Create for leadership and expert-track development | Transfer depends on manager coaching capability; must pair with structured reflection |
+| Coaching / mentoring | Evaluate–Create; individual development plans | Does not scale; requires trained coaches |
+
+Blended designs must specify the cognitive role of each modality: online for
+declarative acquisition, synchronous for practice and feedback, community for
+sustained transfer. Do not use blended as a synonym for "some online and some
+classroom."
+
+## Assessment Design
+
+Distinguish formative from summative before designing any assessment item:
+
+- **Formative** — occurs during learning; purpose is to provide feedback
+  and adjust instruction. Low or zero stakes. Examples: knowledge checks,
+  scenario prompts with immediate feedback, reflection prompts.
+- **Summative** — occurs at module or programme end; purpose is to certify
+  achievement of objectives. Stakes may be high (certification, compliance
+  sign-off). Examples: scored knowledge test, observed performance assessment,
+  portfolio artefact.
+
+Multiple-choice item quality criteria:
+
+1. Stem is a complete question or problem, not an incomplete sentence.
+2. One unambiguously correct answer; distractors are plausible but
+   definitively wrong to a subject-matter expert.
+3. Distractors represent common misconceptions or procedural errors, not
+   random wrong answers.
+4. No "all of the above" or "none of the above" options.
+5. Item tests the objective's stated Bloom level — not a lower level
+   (recall substitute for an Apply objective).
+
+Performance-based assessments are required when the objective is at Apply
+level or above and the job consequence of failure is high (safety, compliance,
+revenue). A written test is not a substitute for an observed simulation or
+work sample when the objective demands execution.
+
+Never assess content that was not taught. Assessment items must have
+traceable alignment to a learning objective and to a learning activity.
+
+## Transfer-of-Learning Measurement
+
+Use the Kirkpatrick four-level model as a minimum evaluation framework:
+
+| Level | Question | Minimum method |
+|---|---|---|
+| 1 — Reaction | Did learners find it relevant and well-executed? | End-of-course survey with NPS + open comment |
+| 2 — Learning | Did learners acquire the intended knowledge or skill? | Objective-aligned assessment with pre/post comparison where feasible |
+| 3 — Behaviour | Are learners applying the learning on the job? | Structured observation, manager checklist, or work-sample audit at 30/60/90 days post-training |
+| 4 — Results | Did the business outcome move? | Business metric comparison (defined before training) |
+
+Level 1 (smile-sheet) is not evidence of learning. Report it, but never
+use it as the primary evidence of programme value to a business stakeholder.
+Level 3 measurement is mandatory for any programme with a per-learner cost
+above a threshold set in the programme plan, or for safety, compliance, or
+critical-path skill programmes regardless of cost.
+
+Phillips ROI extends Level 4 by isolating the training contribution (using
+control group, trend analysis, or expert estimate), converting outcomes to
+monetary value, and comparing to programme cost. Apply Phillips ROI only
+when the business outcome is directly monetisable and the attribution
+methodology can withstand scrutiny; never fabricate an isolation factor.
+
+## Accessibility and Inclusion
+
+Digital learning materials must meet WCAG 2.2 AA as a minimum:
+
+- Video content requires accurate captions; audio description is required
+  when visual content carries meaning not conveyed in the audio track.
+- All interactive elements are keyboard-navigable and have accessible names.
+- Colour is not the sole means of conveying information (error states,
+  status indicators, charts).
+- Text has a contrast ratio of at least 4.5:1 for normal text, 3:1 for
+  large text.
+
+Apply Universal Design for Learning (UDL) principles:
+
+- Provide multiple means of representation (text + visual + audio).
+- Provide multiple means of action and expression (write, speak, draw,
+  demonstrate).
+- Provide multiple means of engagement (choice, relevance framing,
+  self-assessment scaffolding).
+
+Cognitive load management: decompose complex procedures into no more than
+seven ± two chunks per screen or segment; separate intrinsic load (content
+complexity) from extraneous load (navigation, formatting, decorative
+elements) by removing the latter.
+
+## Anti-patterns
+
+| Anti-pattern | Why it fails | Correct approach |
+|---|---|---|
+| Training as default fix | Sends learners through instruction when the gap is motivational or environmental; wastes budget and credibility | Complete a five-source needs diagnosis before committing to training |
+| Smile-sheet as proof of effectiveness | Level 1 satisfaction does not correlate with Level 2 learning or Level 3 behaviour change | Define Level 3 measurement criteria before the programme launches |
+| Dump-and-pray content density | Overloaded slides and long-form e-learning; learners do not retain dense information transfer | Apply the minimum viable content rule: include only what changes the target behaviour |
+| No transfer plan | Programme ends at delivery; no follow-up, no manager briefing, no job aid | Design transfer supports (job aids, manager talking points, 30-day check-in) as deliverables in the programme scope |
+| Accessibility as afterthought | Late-stage remediation is 3–10× more expensive than designing accessibly from the start; legal exposure | Include WCAG 2.2 AA compliance as a Definition of Done on all digital assets |
+| Vague learning objectives | "Understand the policy" cannot be assessed and gives instructional designers no design target | Author every objective to the observable + measurable + levelled standard before design begins |
+| Modality as status signal | Classroom training perceived as more serious; virtual/async perceived as lower quality | Select modality by objective–audience–budget fit, not by stakeholder prestige associations |
+
+## Cross-References
+
+- `core/technical-writing` — for documentation deliverables (job aids,
+  facilitator guides, learner workbooks) that accompany instructional
+  programmes.
+- `frontend/accessibility-and-wcag` — for WCAG 2.2 AA implementation
+  detail in digital courseware and LMS integrations.
+- `core/code-review-checklist` — when reviewing or commissioning custom
+  LMS or courseware-authoring tooling.
+
+## ADR Anchors
+
+- **ADR-058** — governance on domain skill introduction; this file is a
+  seed-only domain skill in the `training-l-and-d` bucket.

package/.claude/skills/domains/voice-ai/skills/voice-ai-integration/SKILL.md

@@ -0,0 +1,405 @@
+---
+name: voice-ai-integration
+description: >
+  Production discipline for voice AI integration covering ASR provider
+  selection (Whisper-large / AssemblyAI / Deepgram / Speechmatics /
+  Google STT) by accuracy, latency, pricing, and language coverage; TTS
+  provider selection (ElevenLabs / OpenAI / Cartesia / Azure Neural /
+  Google Cloud TTS) with latency-to-first-byte and emotional control
+  tradeoffs; real-time streaming via WebRTC and WebSocket with partial
+  transcription and jitter buffer configuration; speaker diarization with
+  error rate budgets; end-to-end latency budgets anchored to hard numbers;
+  conversational state management with barge-in and VAD; fallback handling
+  across provider degradation events; and PII redaction, consent recording,
+  and LGPD/GDPR compliance per jurisdiction. Use when: designing or
+  reviewing a voice pipeline, selecting ASR or TTS providers, defining
+  latency SLAs, architecting real-time audio streaming, or auditing
+  transcript privacy and retention controls.
+owner: Rafael Drummond (Voice AI Integration Engineer, domain persona)
+tier: domain:voice-ai
+scope_tags: [voice-ai, asr, tts, diarization, real-time-audio, webrtc, latency-budget]
+inspired_by:
+  - source: msitarzewski/agency-agents/engineering/engineering-voice-ai-integration-engineer.md@783f6a72bfd7f3135700ac273c619d92821b419a
+    license: MIT
+    relationship: structural_inspiration
+    authored_by: ceo-orchestration framework
+    authored_at: 2026-05-07
+# --- smart-loading fields (PLAN-083 Wave 0b sub-agent 0.7c) ---
+domain: voice-ai
+priority: 8
+risk_class: low
+stack: []
+context_budget_tokens: 500
+inactive_but_retained: true
+repo_profile_binding:
+  frontend: {active: false, priority: 10}
+  engine: {active: false, priority: 10}
+  fintech: {active: false, priority: 10}
+  trading-readonly: {active: false, priority: 10}
+  generic: {active: false, priority: 10}
+activation_triggers: []
+# --- K1 paths: native file-touch activation (PLAN-135 W3 unit k1a) ---
+paths:
+  - "**/voice/**"
+  - "**/asr/**"
+  - "**/tts/**"
+  - "**/webrtc/**"
+  - "**/diarization/**"
+---
+
+# Voice AI Integration
+
+## Cardinal Rule
+
+Latency and accuracy are measured, not assumed. Every voice pipeline
+component carries a hard number — end-to-end conversational response
+must remain below 800 ms, ASR partial results within 200 ms of speech
+offset, synthesis time-to-first-byte within 200 ms of inference output.
+Any component claiming acceptable latency without a benchmark run under
+realistic load is not acceptable. "Feels fine" is not an SLA. Record
+the measurement methodology alongside the number so regressions can be
+detected in CI before reaching production.
+
+---
+
+## Fail-Fast Rule
+
+A voice pipeline MUST NOT enter production without satisfying the
+following four gates: (1) WER is measured against a domain-specific
+holdout set — not the provider's published general-language benchmark;
+(2) end-to-end latency has been measured at p95 under concurrent load,
+not just median in isolation; (3) PII redaction runs as a named,
+independently testable pipeline stage before any transcript is written
+to persistent storage or delivered downstream; (4) provider failover
+has been exercised by injecting a provider error and confirming
+graceful degradation rather than silent failure. If any gate is
+unverified, the pipeline is not production-ready.
+
+---
+
+## When to Apply
+
+Activate this skill when the work involves any of the following:
+
+- Selecting an ASR provider for a new or migrated voice pipeline.
+- Selecting a TTS provider for speech synthesis in a conversational or
+  broadcast context.
+- Designing or reviewing real-time audio streaming architecture via
+  WebRTC or WebSocket.
+- Defining latency budgets or SLAs for a conversational AI product.
+- Implementing or auditing speaker diarization in a multi-participant
+  recording pipeline.
+- Adding barge-in, interruption recovery, or full-duplex audio to a
+  voice interface.
+- Designing fallback and graceful degradation paths for provider
+  outage or confidence-threshold failures.
+- Reviewing transcript privacy controls, consent recording, or
+  retention policies under LGPD, GDPR, or wiretap statutes.
+
+Skip when: the task is pure NLP over pre-existing text transcripts
+(no audio path); the integration is a single-call transcription batch
+job with no latency requirement and no PII exposure; or the context
+is media subtitle generation for a pre-recorded non-interactive asset
+(use a batch transcription pattern instead).
+
+---
+
+## ASR Provider Selection
+
+Provider selection is a function of four independent axes. Choose one
+axis as the binding constraint before evaluating the others.
+
+Select one axis as the binding constraint before evaluating the others.
+
+WER claims from providers are measured on clean studio benchmarks. Every
+deployment must benchmark WER on a domain-specific holdout set — at minimum
+60 utterances representing actual speakers, vocabulary, accents, and
+recording conditions of the target use case.
+
+| Provider | WER (clean) | Streaming | Notes |
+|---|---|---|---|
+| Whisper large-v3 (local) | 3–7% | No native stream | GPU required; no network RTT |
+| Deepgram Nova-2 | 5–9% | Yes, stable partials | Strong telephony tuning |
+| AssemblyAI Universal-2 | 6–10% | Yes, partials revise | Diarization bundled |
+| Speechmatics | 7–12% | Yes | Best multilingual coverage |
+| Google STT v2 | 8–14% | Yes, interim results | Wide language coverage |
+
+For accented, domain-specific, or noisy audio these rankings shift
+materially. Benchmark on actual production audio before committing.
+
+Pricing: all cloud providers charge per audio-minute; rates shift at volume
+tiers. Hybrid routing (local Whisper for sensitive or high-volume workloads;
+cloud for burst or language gaps) is valid when routing complexity is
+justified by volume.
+
+Language: Speechmatics provides the broadest certified multilingual coverage.
+Whisper large-v3 supports 99 languages but quality degrades below the top 20
+by training corpus size. Deepgram and AssemblyAI are strong in English and
+Spanish; limited certification elsewhere. Benchmark WER per language
+independently for multilingual deployments.
+
+---
+
+## TTS Provider Selection
+
+TTS selection depends on voice identity requirements, latency budget,
+and whether emotional or prosodic control is required.
+
+| Provider | TTFB (ms) | Cloning | Notes |
+|---|---|---|---|
+| Cartesia | 80–200 | Yes | Lowest TTFB; streaming-first |
+| ElevenLabs | 100–300 | Yes (custom voice) | Highest naturalness; consent required |
+| Azure Neural TTS | 150–350 | Yes | Enterprise SLA; SSML emotional control |
+| Google Cloud TTS | 150–400 | No | Wide language coverage |
+| OpenAI TTS | 200–400 | No | Simple API; preset voices |
+
+TTFB target: synthesized speech must begin playing within 200 ms of
+inference completion. Providers with median TTFB above 400 ms under load
+require buffering strategies that degrade conversational responsiveness.
+
+Voice cloning requires explicit consent from the voice subject. Retain
+consent artifacts for the lifetime of any cloned voice asset. Synthesizing
+a recognizable voice without consent creates legal exposure in most
+jurisdictions.
+
+---
+
+## Real-Time Streaming
+
+### WebRTC vs WebSocket
+
+WebRTC is the correct choice for browser-to-server audio when
+sub-300 ms round-trip latency is required. It provides built-in
+adaptive bitrate, jitter buffer management, and DTLS-SRTP encryption.
+The operational cost is higher: STUN/TURN server infrastructure,
+ICE negotiation latency on connection establishment, and complexity
+in NAT traversal for mobile clients.
+
+WebSocket is the correct choice when the audio source is a server-side
+process, a mobile SDK with direct control over encoding, or any context
+where WebRTC's ICE negotiation overhead is unacceptable for use-case
+latency. WebSocket requires explicit jitter buffer implementation in
+the application layer.
+
+Do not choose based on familiarity. Choose based on whether the audio
+source is a browser (WebRTC) or a server-side or native SDK (WebSocket).
+
+### Chunked Transcription, Partials, and Jitter Buffer
+
+Send audio in fixed-size chunks (100–250 ms). Below 80 ms, per-request
+overhead grows without proportional latency reduction. Partial results are
+not final — they may revise on the next final event. Do not commit a
+partial result to a database record or downstream action; wait for the
+final-result flag.
+
+For WebSocket pipelines, implement a jitter buffer of 150–300 ms before
+passing audio to the ASR engine. Target depth: two to three times the
+observed p95 packet inter-arrival jitter, measured under load.
+Undersized buffers produce gap artifacts that degrade WER silently.
+
+---
+
+## Speaker Diarization
+
+### Online vs Offline
+
+Online diarization (streaming, low-latency) operates on a rolling
+window and produces approximate speaker boundaries. It degrades on
+cross-talk and requires a speaker-count assumption for best results.
+Use for near-real-time applications where latency matters more than
+boundary precision.
+
+Offline diarization (batch, post-transcription) runs on the full
+recording and produces higher-accuracy boundaries. Use for call
+analytics, meeting transcripts, and compliance recordings where
+precision matters more than latency.
+
+### Speaker Count Assumption
+
+Provide the expected speaker count when known. Diarization models
+that estimate speaker count from audio produce materially higher
+diarization error rates (DER) when the audio contains more than
+four speakers or any cross-talk. If speaker count is unknown and
+cross-talk is expected, budget for a DER of 15–25% and route
+low-confidence segments for human review.
+
+### Cross-Talk and DER Budget
+
+Segments with two or more overlapping speakers cannot be reliably
+attributed. Flag them explicitly in the output schema — do not silently
+assign to whichever speaker model heuristics select.
+
+DER targets (measured on a domain-representative holdout set, not
+provider benchmark corpora):
+
+- Clean single-microphone, low overlap: DER < 5%.
+- Telephony or multi-participant: DER < 12%.
+- High-overlap or noisy: DER < 20%, with manual review routing for
+  flagged segments.
+
+---
+
+## Latency Budgets
+
+End-to-end conversational latency (speech-offset to synthesized-audio
+play-start) must remain below 800 ms for natural conversational feel.
+Allocate the budget across pipeline components:
+
+| Component | Budget | Notes |
+|---|---|---|
+| ASR partial first result | ≤ 200 ms | From speech onset; streaming required |
+| ASR final result | ≤ 400 ms | From end of utterance |
+| Inference / LLM call | ≤ 300 ms | First token; streaming generation |
+| TTS time-to-first-byte | ≤ 200 ms | From inference first-token |
+| Audio delivery to client | ≤ 50 ms | Network; WebRTC/WebSocket buffer |
+| **Total end-to-end** | **≤ 800 ms** | **p95 target under load** |
+
+Budgets are measured at p95 under concurrent load, not median in
+isolation. Overage in any single component must be compensated by
+reducing budget in an adjacent component — the total envelope does
+not flex. If the 800 ms target cannot be met, document the actual
+p95 measurement and the technical constraint preventing compliance.
+Never accept "feels fine."
+
+---
+
+## Conversational State Management
+
+### Turn-Taking and Barge-In
+
+Maintain explicit turn state: LISTENING, PROCESSING, SPEAKING.
+Transition logic must be deterministic — combine VAD silence threshold
+with a configurable end-of-utterance timeout (300–500 ms typical;
+tune to domain speech patterns). Do not rely on silence alone.
+
+Barge-in: the system must stop synthesis and return to LISTENING when
+VAD detects speech onset during a synthesized response. Implement by
+(1) keeping the ASR streaming session active during synthesis, (2)
+triggering synthesis cancellation on VAD onset above threshold, (3)
+discarding buffered TTS audio beyond the cancellation point. Barge-in
+without synthesis cancellation produces overlapping audio — users
+interpret this as a broken product. Do not ship without it.
+
+### Interruption Recovery
+
+When barge-in fires, preserve the generation state (partial LLM response
+at the interruption point, turn context). Log the cut point; resumption
+logic determines whether to continue the interrupted response or discard
+it based on the new user utterance.
+
+---
+
+## Fallback Handling
+
+### Provider Degradation and ASR Confidence
+
+Monitor ASR and TTS provider health via p95 latency and error rate.
+Circuit-breaker: if ASR error rate exceeds 5% in a 60-second rolling
+window, switch to secondary; return to primary after 120 s of
+below-threshold operation. Route provider-switch events to the
+observability pipeline — silent failover is acceptable UX but
+unacceptable operationally.
+
+When ASR returns confidence below threshold (starting point: 0.7 on
+0–1 scale), trigger a clarification prompt rather than proceeding on
+uncertain input. In high-stakes contexts (medical, financial, legal),
+do not silently proceed on low-confidence segments. Configure the
+threshold per domain.
+
+### TTS Failover and Offline Impossibility
+
+Pre-configure a failover voice and validate it is intelligible in the
+deployment language before it is needed. Do not dynamically select an
+untested voice on failover.
+
+Cloud-dependent voice pipelines cannot degrade to a fully offline mode —
+the capability disappears with the provider. Design for graceful
+degradation: surface a clear message when no provider is reachable;
+preserve conversation context so the session can resume when connectivity
+returns.
+
+---
+
+## Privacy and Compliance
+
+### PII Redaction
+
+PII redaction must run as a named, independently testable pipeline
+stage before any transcript is written to persistent storage or
+delivered to a downstream consumer. Redaction must be configured, not
+assumed. Test the redaction stage independently with known PII fixtures:
+phone numbers, CPF/CNPJ (Brazil), SSN patterns, credit card numbers,
+names in context.
+
+Do not treat redaction as an output filter. Run it as an intermediate
+stage so that confidence scores on redacted segments are preserved for
+audit — the model output is immutable, the stored transcript is redacted.
+
+### Consent Recording per Jurisdiction
+
+- Brazil (LGPD): informed consent required; purpose specified at collection;
+  consent artifact retained for the lifetime of the recording.
+- EU/EEA (GDPR): lawful basis required; data subject right to erasure;
+  cross-border transfer requires adequacy decision or SCCs.
+- United States: wiretap law varies by state; California, Connecticut, and
+  Maryland require all-party consent; federal one-party consent does not
+  supersede stricter state law.
+
+Consent capture and artifact storage are first-class pipeline components.
+
+### Storage Retention
+
+Define retention windows per data class: raw audio (30-day default unless
+a specific requirement extends), redacted transcript (longer acceptable,
+linked to consent artifact), speaker embeddings (biometric data under LGPD
+and GDPR; explicit consent required; shorter windows). Automated deletion
+must be implemented and tested — transcripts beyond policy window create
+compliance liability regardless of access history.
+
+---
+
+## Anti-Patterns
+
+| Anti-Pattern | Why It Fails |
+|---|---|
+| Skipping jitter buffer configuration | Network variability creates audio gaps that degrade ASR WER without any error signal; failure is silent |
+| Accepting provider WER claim without domain benchmark | Published benchmarks use clean studio audio; production WER on telephony or multi-accent audio is typically 2–4× higher |
+| Single-provider dependency with no failover | Provider outages are not rare; a voice pipeline with no secondary provider has zero availability during provider incidents |
+| Missing barge-in implementation | Users interpret overlapping synthesized speech as a broken product; full-duplex without barge-in is not shippable |
+| PII redaction as output filter only | If redaction fails silently, unredacted PII reaches persistent storage; redaction must be a verifiable intermediate stage with test coverage |
+| Treating silence as end-of-utterance sole signal | VAD false positives on pauses inside long utterances cause premature turn switches; combine with configurable timeout |
+| Using partial ASR results as committed output | Partials revise on finalization; committing a partial to a downstream action or database record produces incorrect or duplicate data |
+
+---
+
+## Cross-References
+
+- `core/security-and-auth` — authentication patterns for API key management
+  to ASR and TTS provider endpoints; credential rotation discipline.
+- `core/compliance-lgpd` — LGPD data subject rights implementation, consent
+  artifact storage, and purpose limitation for audio and transcript data.
+- `core/architecture-decisions` — ADR lifecycle for provider selection
+  decisions and latency budget commitments that cross architectural
+  boundaries.
+
+---
+
+## ADR Anchors
+
+Decisions affecting provider selection, latency SLAs, PII handling, or
+consent architecture MUST be recorded as ADRs.
+
+- **ADR-058** (Brainstorm Gate and Two-Pass Review) — provider-selection
+  decisions, latency-budget changes, and adapter-pattern code are all
+  subject to two-pass adversarial review before deployment.
+- **Adapter layer pattern (project-internal):** voice provider clients
+  wrapped behind an adapter interface to enable failover and
+  substitution without upstream code changes (project-specific ADR
+  authored at adoption time).
+- **Provider selection ADR:** evaluated alternatives, WER benchmark
+  methodology, binding constraint, decision. Updated on every provider
+  change.
+- **Latency budget ADR:** 800 ms target, per-component allocation,
+  measurement methodology, approved exceptions with stated constraint.
+- **PII and consent ADR:** redaction stage design, retention windows per
+  data class, consent capture mechanism, jurisdiction scope.