sinapse-ai 9.4.0 → 9.5.0

package/squads/squad-content/squad.yaml

@@ -1,14 +1,14 @@
 # ══════════════════════════════════════════════════════════════════════════════
-# SQUAD MANIFEST — squad-content v1.0
+# SQUAD MANIFEST — squad-content v2.0
 # ══════════════════════════════════════════════════════════════════════════════
 # The world-class content intelligence squad for detecting signals, planning
 # editorially, structuring narratives, producing content at scale, adapting
 # natively per platform, auditing quality, and measuring performance.
-# 7 agents | 90 tasks | 6 workflows | 16 KBs | 7 templates | Full coverage
+# 7 agents | 90 tasks | 6 workflows | 32 KBs | 7 templates | Full coverage
 # ══════════════════════════════════════════════════════════════════════════════
 
 name: squad-content
-version: 1.0.0
+version: 2.0.0
 description: >
   Squad de classe mundial para inteligencia de conteudo: deteccao de sinais em
   tempo real, planejamento editorial estrategico, producao de conteudo em escala
@@ -23,16 +23,17 @@ slashPrefix: SINAPSE
 
 metadata:
   created_at: "2026-03-13"
-  updated_at: "2026-03-13"
+  updated_at: "2026-04-12"
   created_by: "squad-creator (Craft)"
+  last_upgraded_by: "Story fw-v2.4 (squad-content v2.0 upgrade) via @sinapse-orqx Imperator"
   quality_target: "5.0/5.0 em todas as dimensoes"
   agents_count: 7
   tasks_count: 90
-  knowledge_bases_count: 16
+  knowledge_bases_count: 32
   checklists_count: 3
   templates_count: 7
   workflows_count: 6
-  total_files: 130
+  total_files: 146
 
 # ══════════════════════════════════════════════════════════════════════════════
 # QUALITY STANDARDS — Regras de Robustez (v1.0)
@@ -281,42 +282,58 @@ components:
     - onboarding-content-cycle.yaml      # Setup de cliente novo: Big Idea → primeiro sprint
 
 # ══════════════════════════════════════════════════════════════════════════════
-# KNOWLEDGE BASES — 14 BASES
+# KNOWLEDGE BASES — 32 BASES (reconciled to disk 2026-04-12, v2.0.0)
 # ══════════════════════════════════════════════════════════════════════════════
 
   knowledge_bases:
+    # Canon playbooks (v2 — story fw-v2.4)
+    - ai-native-content-loop.md              # Canon 6-phase loop: signal→ai draft→human edit→adapt→measure→retrofeed
+    - task-ownership-map.md                  # 90-task ownership matrix, imbalance audit, routing reference
+    - signal-intelligence-v2.md              # 2026 source landscape, classification taxonomy, routing, quality scoring
+
     # Strategy & Planning
-    - content-strategy-frameworks.md     # Pulizzi, Halvorson, Rose, CMI, HubSpot
-    - content-spine-engine.md            # Motor central: tese-mecanismo-prova-direcao
-    - editorial-governance.md            # Governanca, RACI, lifecycle, style guide
+    - content-strategy-master-reference.md   # Master content strategy reference
+    - editorial-strategy-frameworks.md       # Editorial strategy frameworks (Pulizzi, Halvorson, Rose, CMI, HubSpot)
+    - content-architecture-taxonomy.md       # Taxonomy, topic clusters, Rosenfeld & Morville
+    - content-spine-engine.md                # Motor central: tese-mecanismo-prova-direcao
+    - content-governance-principles.md       # Governanca, RACI, lifecycle, style guide
 
     # Narrative & Writing
-    - narrative-structures.md            # Campbell, Miller SB7, Heath SUCCESs, Hook Systems
-    - writing-rules-engine.md            # Regras de escrita por formato, construcao causal
+    - narrative-structures-library.md        # Campbell, Miller SB7, Heath SUCCESs, Hook Systems
+    - hook-formulas-library.md               # Hook Point, pattern interrupts, attention capture formulas
+    - copywriting-formulas-library.md        # Persuasion formulas shared with squad-copy
+    - strategic-content-authority.md         # Processo visivel como ativo de autoridade
+    - ai-leverage-content-framework.md       # Transformar updates de IA em consequencia empresarial
+    - social-compression-framework.md        # Compressao de tese densa sem perder substancia
 
     # Platform & Distribution
-    - platform-algorithms.md             # Algoritmo por plataforma, versionado
-    - template-contract-system.md        # Contratos tecnicos por plataforma x formato
-    - content-atomization-repurposing.md # Atomizacao + matrizes de transformacao
+    - platform-algorithm-intelligence.md     # Algoritmo por plataforma, versionado
+    - social-algorithms-master-reference.md  # Master reference for social algorithms
+    - template-contract-system.md            # Contratos tecnicos por plataforma x formato
+    - content-repurposing-playbook.md        # Atomizacao + matrizes de transformacao
+    - content-formats-encyclopedia.md        # Encyclopedia of content formats per platform
+    - email-newsletter-strategy.md           # Newsletter economics + lifecycle
+    - video-audio-content-playbook.md        # Video/audio content playbook
 
     # Quality & Measurement
-    - eeat-quality-standards.md          # Google E-E-A-T, quality signals, trust
-    - content-scoring-models.md          # Metricas, engagement, predictive, attribution
-    - viral-potential-scoring.md         # SPV unificado pre/pos, STEPPS, Hook Point
+    - content-quality-criteria.md            # Quality criteria for E-E-A-T, voice, narrative
+    - content-measurement-framework.md       # Measurement framework, KPIs, attribution
+    - viral-potential-scoring.md             # SPV unificado pre/pos, STEPPS, Hook Point
 
     # Intelligence & Architecture
-    - signal-detection-methods.md        # Deteccao de sinais, trend classification
-    - information-architecture.md        # Taxonomia, topic clusters, Rosenfeld & Morville
+    - signal-detection-methods.md            # v1 — foundational signal definitions (UNTOUCHED in v2)
+    - competitor-analysis-methods.md         # Competitor content analysis methods
+    - seo-content-deep.md                    # Deep SEO content strategies
+    - content-references-bibliography.md     # Bibliography of world-class references
 
     # AI & Production
-    - ai-content-production.md           # AI-assisted creation, prompt engineering, governance
-
-    # Authority & Process
-    - strategic-content-authority.md     # Processo visivel como ativo de autoridade — bastidor, criterio e prova de competencia
+    - ai-content-production.md               # AI-assisted creation, prompt engineering, governance
+    - content-tech-stack.md                  # Content ops tech stack
 
-    # AI & Leverage
-    - ai-leverage-content-framework.md   # Transformar updates de IA em conteudo de consequencia empresarial
-    - social-compression-framework.md    # Compressao de tese densa em conteudo social-native sem perder substancia
+    # Context & Routing
+    - brazilian-content-context.md           # pt-BR cultural + linguistic context
+    - cross-squad-integration-protocol.md    # Cross-squad handoff protocol
+    - routing-catalog.md                     # Intra-squad routing catalog
 
   checklists:
     - content-quality-checklist.md       # QA pre-publicacao: voice, narrative, E-E-A-T, platform
@@ -403,10 +420,153 @@ cross_squad_connections:
     - squad-copy: "Tecnicas de persuasao, frameworks de copy, awareness levels"
     - squad-growth: "SEO keywords, pillar-cluster data, search intent mapping"
 
+# ══════════════════════════════════════════════════════════════════════════════
+# AGENT TASK OWNERSHIP — v2.0 backfill (source of truth: task-ownership-map.md KB)
+# ══════════════════════════════════════════════════════════════════════════════
+# This block makes task ownership deterministic. Every task has exactly one
+# primary agent. See task-ownership-map.md for secondary mappings and rationale.
+# Added in v2.0 (Story fw-v2.4) — AC2 of the rebalance story.
+# ══════════════════════════════════════════════════════════════════════════════
+
+agent_task_ownership:
+  content-orqx:
+    primary_count: 5
+    phase_coverage: "cross-phase (coordinator)"
+    tasks:
+      - orchestrate-content-pipeline.md
+      - manage-content-sprint.md
+      - coordinate-cross-squad.md
+      - triage-content-requests.md
+      - triage-urgent-signal.md
+  signal-intelligence:
+    primary_count: 10
+    phase_coverage: "phase 1 (signal)"
+    tasks:
+      - scan-daily-signals.md
+      - detect-trending-topics.md
+      - map-cultural-signals.md
+      - curate-weekly-briefing.md
+      - monitor-competitor-content.md
+      - alert-opportunity-windows.md
+      - classify-signal-temperature.md
+      - map-signal-to-pillar.md
+      - configure-signal-sources.md
+      - curate-ugc-signals.md
+  editorial-strategist:
+    primary_count: 10
+    phase_coverage: "pre-phase-1 + phase 6"
+    tasks:
+      - define-big-idea.md
+      - define-editorial-pillars.md
+      - create-editorial-calendar.md
+      - plan-content-sprint.md
+      - align-content-to-funnel.md
+      - map-content-gaps.md
+      - plan-seasonal-content.md
+      - prioritize-themes.md
+      - balance-content-mix.md
+      - define-content-kpis.md
+  content-engineer:
+    primary_count: 25
+    phase_coverage: "phase 2 (ai draft)"
+    load_status: "OVER-LOADED — documented, split candidate for future story"
+    tasks:
+      - design-content-spine.md
+      - design-narrative-arc.md
+      - structure-carousel-progression.md
+      - structure-video-script.md
+      - design-content-series.md
+      - apply-storybrand-framework.md
+      - design-pillar-cluster-model.md
+      - create-content-brief.md
+      - design-hook-system.md
+      - structure-argument-progression.md
+      - define-content-depth-levels.md
+      - write-carousel-content.md
+      - write-instagram-caption.md
+      - write-linkedin-post.md
+      - write-blog-article.md
+      - write-video-script.md
+      - write-reel-script.md
+      - write-newsletter-editorial.md
+      - write-thread-content.md
+      - write-opinion-post.md
+      - write-case-study.md
+      - write-educational-content.md
+      - write-product-content.md
+      - verify-algorithmic-hooks.md
+      - batch-content-production.md
+  platform-specialist:
+    primary_count: 17
+    phase_coverage: "phase 4 (platform adapt)"
+    tasks:
+      - adapt-for-instagram-feed.md
+      - adapt-for-instagram-stories.md
+      - adapt-for-instagram-reels.md
+      - adapt-for-linkedin.md
+      - adapt-for-blog-seo.md
+      - adapt-for-tiktok.md
+      - adapt-for-twitter-threads.md
+      - define-posting-schedule.md
+      - create-platform-guidelines.md
+      - optimize-for-algorithm.md
+      - design-content-atomization.md
+      - register-template-contract.md
+      - update-template-specs.md
+      - adapt-content-language.md
+      - batch-platform-adaptation.md
+      - flag-content-for-amplification.md
+      - integrate-brand-visual-assets.md
+  content-governor:
+    primary_count: 12
+    phase_coverage: "phase 3 (human edit) + phase 6 (governance)"
+    tasks:
+      - audit-content-library.md
+      - check-brand-consistency.md
+      - validate-content-quality.md
+      - manage-content-taxonomy.md
+      - organize-content-archive.md
+      - create-content-governance.md
+      - define-content-standards.md
+      - audit-template-compliance.md
+      - validate-content-compliance.md
+      - process-content-rejection.md
+      - import-existing-content-library.md
+      - update-content-preferences.md
+  content-analyst:
+    primary_count: 11
+    phase_coverage: "phase 5 (measure) + phase 6 (retrofeed)"
+    tasks:
+      - analyze-content-performance.md
+      - score-content-quality.md
+      - measure-content-roi.md
+      - track-engagement-patterns.md
+      - benchmark-against-competitors.md
+      - evaluate-eeat-compliance.md
+      - generate-content-report.md
+      - identify-top-performers.md
+      - predict-content-performance.md
+      - analyze-audience-behavior.md
+      - generate-retrofeed-insights.md
+
 # ══════════════════════════════════════════════════════════════════════════════
 # CHANGELOG
 # ══════════════════════════════════════════════════════════════════════════════
 #
+# v2.0.0 (2026-04-12) — Content Loop Canon + Task Ownership + Signal v2
+#   Story: fw-v2.4 (squad-content agent rebalance + AI-native loop)
+#   Epic: EPIC-framework-upgrade-v2
+#   Changes:
+#     - NEW KB: ai-native-content-loop.md — canon 6-phase loop playbook
+#     - NEW KB: task-ownership-map.md — 90-task ownership matrix with imbalance audit
+#     - NEW KB: signal-intelligence-v2.md — 2026 source landscape, classification, quality scoring
+#     - YAML reconciled to disk reality: knowledge_bases_count 16 → 32 (16 prior undeclared)
+#     - YAML gained agent_task_ownership block (authoritative primary mapping)
+#     - total_files 130 → 146
+#     - Zero edits to the 7 agent persona files (AC5 respected)
+#     - Zero edits to signal-detection-methods.md v1 (v2 is additive companion)
+#   Imbalance audit logged: content-engineer over-loaded (25 primaries) — split candidate for future story
+#
 # v1.0.0 (2026-03-13) — Initial Release
 #   Squad criada com 3 rodadas de provocacao + micro-rodada cirurgica
 #   7 agentes, 90 tasks, 6 workflows, 14 KBs, 3 checklists, 7 templates

package/squads/squad-copy/knowledge-base/ai-copy-human-loop-canon.md

@@ -0,0 +1,235 @@
+# Knowledge Base: AI Copy Human-in-the-Loop Canon
+
+> **Status:** Canon — canonical source of truth for when AI drafts ship, when humans intervene, and how the handoff works inside `squad-copy` (Quill).
+> **Version:** 1.0 (shipped with squad-copy v2.0.0 framework-upgrade / fw-v2.5)
+> **Raises to canon:** `ai-copy-production.md` (capability note, still valid as reference)
+> **Depends on:** `market-awareness-spectrum.md`, `persuasion-architecture-framework.md`, `voice-and-tone-system.md`
+
+## Why this doc exists
+
+`ai-copy-production.md` documents **capability** — what AI does well, where humans are insubstituible. It is correct and still useful, but it is not a canon — it does not tell an operator "on this specific copy job, should you ship the AI draft, light-edit it, rewrite it, or scrap it and start over?" That decision question is the single most common one in day-to-day copy work in 2026, and getting it wrong in either direction costs money — over-editing wastes hours on copy AI could have drafted well, under-editing ships commodity copy that underperforms.
+
+This canon is the decision framework. Every copy agent in `squad-copy` that uses AI assistance MUST run the 4-gate decision tree before deciding the edit depth on a piece of copy.
+
+---
+
+## 1. The 4-gate decision tree
+
+Given an AI draft, the operator asks FOUR questions in order. The first **NO** answer determines the action.
+
+```
+Gate 1: TRUST — Can the AI draft ship as-is with spot-check only?
+   YES → Ship it (proofread + fact check only)
+   NO  → Gate 2
+
+Gate 2: LIGHT EDIT — Can a 10-20% touch-up make it ship-ready?
+   YES → Apply light edit (voice, polish, hooks)
+   NO  → Gate 3
+
+Gate 3: HEAVY REWRITE — Is the structure usable as scaffolding?
+   YES → Heavy rewrite on top of AI scaffolding
+   NO  → Gate 4
+
+Gate 4: SCRAP — Is there any salvageable angle or data?
+   YES → Extract what's salvageable, start over from brief
+   NO  → Scrap entirely, write from zero
+```
+
+Each gate has its own signal set. The signals are ordered from highest-leverage to lowest — any strong NO on a high-leverage signal drops the draft to the next gate immediately.
+
+---
+
+## 2. Gate 1 — Trust
+
+**Ship the AI draft with only proofreading and fact-checking.**
+
+### When Gate 1 passes
+
+All of the following are true:
+- **Stakes are low** — social media post, internal comms, minor UI microcopy, FAQ entry
+- **Market awareness** is informed or aware (per `market-awareness-spectrum.md` Schwartz levels 3-5), i.e. the reader already knows the product/category and needs information, not persuasion
+- **Brand voice is generic-professional** — the brand's voice is not a differentiator, functional clarity is the goal
+- **No claims** — the draft makes no factual/statistical/outcome claims that need verification
+- **Format is standardized** — headline/body/CTA or similar well-trodden pattern
+
+### What Gate 1 copy looks like
+- Product feature bullets, "you can also do X" microcopy, 404 pages, confirmation toasts
+- Social media slot fillers (not hero posts)
+- Internal comms drafts
+- FAQ answers for operational questions
+
+### Required steps before ship
+1. Proofread (human eye, not AI spell-check)
+2. Fact-check every claim
+3. Compliance scan (no regulated claims — health, financial, legal)
+4. Brand voice sanity check — single pass for "does this sound like us"
+
+### Recommended AI driver
+See §6 prompt library — use the `gate-1-draft` template.
+
+---
+
+## 3. Gate 2 — Light edit (10-20% touch-up)
+
+**AI draft is the baseline, human adds voice + polish + conversion levers.**
+
+### When Gate 2 passes (and Gate 1 fails)
+
+- **Stakes are moderate** — email body, landing page section, ad copy variant, blog intro
+- **Market awareness** is solution-aware (Schwartz level 3), reader knows the problem but not your solution
+- **Brand voice is distinctive** but the AI can approximate it with strong prompting
+- **Claims exist** but are verifiable and simple
+- **The hook/CTA needs sharpening** but the body is fine
+
+### What the light edit typically changes
+- **Hook** — replace or sharpen the opening 1-2 sentences (AI defaults are generic)
+- **Voice injection** — swap 5-15% of the vocabulary for brand-specific register (lexicon, metaphors, rhythm)
+- **Conversion lever** — add 1-2 persuasion triggers from `persuasion-triggers-catalog.md` the AI missed
+- **CTA** — rewrite from AI's generic "Learn more" / "Saiba mais" to the brand's canonical CTA pattern
+- **Proof** — inject specific proof elements (numbers, testimonials, case studies) AI did not have
+- **Cut** — remove 10-20% of filler the AI added
+
+### What Gate 2 does NOT do
+- Does not rewrite the structure — if the structure is wrong, escalate to Gate 3
+- Does not add a new angle — if the angle is off, escalate to Gate 3
+- Does not fix factual errors by "softening" them — always correct or escalate
+
+---
+
+## 4. Gate 3 — Heavy rewrite
+
+**AI draft is scaffolding, human authors 50-80% of the final text on top of it.**
+
+### When Gate 3 passes (and Gate 2 fails)
+
+- **Stakes are high** — sales page hero, launch email, ad campaign flagship, VSL script
+- **Market awareness** is product-aware or most-aware (Schwartz levels 4-5), reader needs nuanced persuasion
+- **Brand voice is highly distinctive** and AI approximations lose the specific rhythm, references, or controversy the brand owns
+- **Opinion / point-of-view is required** — AI defaults to consensus, brand defaults to a sharp POV
+- **Complex proof** — multiple studies, custom data, interlocked testimonials
+- **Objection handling is layered** — 3+ objections handled in a specific order with specific reframes
+
+### What the heavy rewrite typically does
+- **Keeps the AI's structural outline** (sections, order, rough length)
+- **Keeps some AI bullets/data points** that are useful (if any)
+- **Replaces 50-80% of the prose** with human-authored voice
+- **Rewrites every section opener** — AI openers are almost always generic
+- **Reorders for conversion** — AI tends to information-order; humans convert-order
+- **Adds the brand's signature moves** — specific metaphors, signature phrases, callbacks
+- **Proof injection at full depth** — not just "we helped N customers" but the specific story
+
+### AI role in Gate 3
+Gate 3 is NOT "AI is useless". AI is still the scaffolding — the outline, the brainstorm, the objection enumeration, the headline variants (pick 1 from 25). The heavy lift is human because the brand signature and POV are non-negotiable on high-stakes copy.
+
+---
+
+## 5. Gate 4 — Scrap and start from zero
+
+**AI draft is not usable. Start from the original brief.**
+
+### When Gate 4 passes (and Gate 3 fails)
+
+- **The angle is wrong** — AI picked a generic angle when the brief needed a contrarian one
+- **The brand voice is so specific** that AI cannot approximate it without producing cringe (Seth Godin-style blogs, David Ogilvy direct-response longform, Brazilian direct-response with specific regional flavor)
+- **Legal/compliance is delicate** — regulated industry, claim scrutiny, political/cultural sensitivity
+- **The piece requires lived experience** — "I built X and here's what I learned" copy where authenticity IS the value prop
+- **The AI draft shows hallucinations** — made-up statistics, fabricated testimonials, invented product features — at which point the draft is contaminated and cannot be "fixed"
+
+### How to use AI when scrapping to start over
+1. Generate 20-50 **hooks/angles** with AI (cheap brainstorming)
+2. Human picks 1 angle
+3. Human writes the piece from zero
+4. AI does **editing passes** on the human draft (copy-editing, not re-drafting)
+5. AI does **variant generation** (headline variants, CTA variants, subject line variants — 10-25 options per slot)
+6. Human picks finalists, ships
+
+The AI is a **research/brainstorm/variant tool** in Gate 4, not a drafter.
+
+---
+
+## 6. AI prompt library (agent-owned)
+
+Every copy agent in `squad-copy` owns 1-2 AI prompt templates aligned to their specialty. The templates live in the agent file and are referenced here for the canonical set. Agent ownership prevents drift and ensures accountability.
+
+| Prompt template | Purpose | Owner agent | Typical gate |
+|----------------|---------|-------------|-------------|
+| `gate-1-draft` | Generic AI draft for low-stakes copy | copy-editor | Gate 1 |
+| `hook-brainstorm` | 25 hook/angle variations for an offer | headline-specialist | Gate 3-4 |
+| `headline-variants` | 25 headline variations from a single angle | headline-specialist | Any |
+| `objection-enumeration` | Enumerate objections for a given audience + offer | copy-strategist | Gate 2-4 |
+| `objection-handler` | Reframe 1 objection in 3 angles | copy-strategist | Gate 2-3 |
+| `proof-synthesis` | Assemble proof elements (testimonials, data, case studies) into persuasive summary | copy-strategist | Gate 2-3 |
+| `cta-variation` | 10 CTA variations (action verb × reward × urgency) | ad-copywriter | Any |
+| `subject-line-cascade` | Cold → warm → hot subject lines for an email sequence | email-copywriter | Gate 1-2 |
+| `vsl-beat-sheet` | VSL structure beat-by-beat from a brief | vsl-scriptwriter | Gate 3 |
+| `social-compression` | Compress long-form into platform-native short form | social-copywriter | Gate 1-2 |
+| `market-awareness-diagnostic` | Diagnose reader's Schwartz awareness level from a brief | copy-strategist | Pre-draft |
+| `voice-audit` | Audit a draft against brand voice canon, flag drift | copy-editor | Any gate's exit check |
+
+**Rule:** Before a prompt template is added to the library, the owner agent must have run the template on ≥ 3 real copy jobs and the quality must be documented in the agent's MEMORY.md. Unproven prompts do not enter the canon library.
+
+---
+
+## 7. Signals cheat sheet — which signals push which way
+
+| Signal | Pushes toward |
+|--------|--------------|
+| Low stakes, generic voice, informed audience | Gate 1 |
+| Moderate stakes, distinctive voice, solution-aware audience | Gate 2 |
+| High stakes, sharp POV, most-aware audience | Gate 3 |
+| Regulated industry / legal sensitivity | Gate 3 or 4 |
+| Requires lived experience / authenticity | Gate 4 |
+| AI hallucinated facts | Gate 4 (draft contaminated) |
+| Complex interlocked proof | Gate 3 |
+| Simple verifiable claims | Gate 1 or 2 |
+| Brand has signature moves (metaphors, callbacks, controversy) | Gate 3 |
+| Copy is template-fillable (FAQ, toast, microcopy) | Gate 1 |
+| Multi-step objection handling required | Gate 3 |
+| Single objection, standard reframe | Gate 2 |
+
+---
+
+## 8. Production hygiene (every AI-assisted piece)
+
+Every AI-assisted copy deliverable MUST record:
+
+1. **Gate used** — which of the 4 gates (Trust / Light / Heavy / Scrap)
+2. **Prompt template(s) used** — from the library in §6
+3. **Model used** — GPT-5, Claude Opus 4.6, Gemini 2.5 Pro, etc.
+4. **Final edit depth** — rough % AI vs human in the shipped version
+5. **Conversion outcome** (if tracked) — linked to the attribution system
+
+This record belongs in the copy job's handoff artifact. The record is used to:
+- Audit the gate decisions over time (are we systematically over/under-editing?)
+- Train operators on gate selection
+- Feed back into the prompt library (which prompts outperform, which drift)
+
+---
+
+## 9. Anti-patterns (forbidden)
+
+- **Skipping gate selection** — shipping whatever AI produced without running the 4-gate decision
+- **Systematic Gate 1** — defaulting to "AI is good enough" for high-stakes copy
+- **Systematic Gate 4** — defaulting to "rewrite everything from zero" for low-stakes copy (wastes the AI leverage)
+- **Editing hallucinated facts instead of flagging them** — a draft with hallucinations is contaminated, escalate to Gate 4 or rerun the prompt with facts in context
+- **Using "in the voice of [living copywriter]"** — legally risky and produces pastiche; use attribute descriptions instead
+- **Shipping without production hygiene (§8)** — if you can't reconstruct the gate + prompt + model used, the job is undocumented
+- **Mixing AI prose and human prose without re-voice-pass** — AI sections stand out from human sections unless a final voice pass unifies them
+- **Treating AI variants as consensus output** — always pick ONE finalist per variant generation, never ship "an average"
+- **Re-generating endlessly to avoid human decision** — if you're on your 10th generation and still unhappy, escalate to Gate 3 or Gate 4 instead of regenerating
+
+---
+
+## 10. Integration with existing KBs
+
+- `ai-copy-production.md` — capability reference (what AI does well / where humans are insubstituible). Still valid; this canon builds on it.
+- `market-awareness-spectrum.md` — Gate 1-4 signals reference the Schwartz awareness levels directly.
+- `persuasion-architecture-framework.md` — the "conversion levers" that Gate 2 light-edits inject.
+- `persuasion-triggers-catalog.md` — individual triggers the Gate 2 edit inserts.
+- `voice-and-tone-system.md` — the voice canon that the Gate 2 voice injection and Gate 3 rewrite enforce.
+- `copywriting-formulas-encyclopedia.md` — structural scaffoldings that AI drafts rarely get right without explicit instruction.
+- `legendary-copywriters-frameworks.md` — the POV/signature moves that Gate 3 rewrites restore.
+
+## Change log
+
+- **2026-04-12 (v1.0)** — Created as part of squad-copy v2.0.0 / fw-v2.5. Raises `ai-copy-production.md` from capability reference to a canonical 4-gate decision framework covering Trust / Light / Heavy / Scrap with signals, AI prompt library (12 templates mapped to owner agents), production hygiene, integration with existing KBs, and anti-patterns.

package/squads/squad-copy/squad.yaml

@@ -1,5 +1,5 @@
 name: squad-copy
-version: "1.0.0"
+version: "2.0.0"
 slashPrefix: SINAPSE
 author: Caio Imori
 license: UNLICENSED
@@ -10,17 +10,32 @@ description: >
 
 metadata:
   created: "2026-03-12"
+  last_updated: "2026-04-12"
   author: "Sinapse"
   tier: 1
-  total_files: 113
-  agents: 12
+  total_files: 135
+  agents: 14
   tasks: 81
   workflows: 6
-  knowledge_bases: 14
+  knowledge_bases: 24
   checklists: 3
   templates: 7
   preferences: 1
 
+changelog:
+  - version: "2.0.0"
+    date: "2026-04-12"
+    story: fw-v2.5
+    epic: EPIC-framework-upgrade-v2
+    highlights:
+      - "Shipped ai-copy-human-loop-canon.md KB — 4-gate decision framework (Trust/Light/Heavy/Scrap) for when AI drafts ship vs when humans intervene. Covers signals per gate, AI prompt library (12 templates mapped to owner agents), production hygiene, integration with existing KBs, anti-patterns."
+      - "Raises ai-copy-production.md from capability reference to canonical decision framework."
+      - "Reconciled yaml counts to disk reality (agents 12->14, knowledge_bases 14->24, total_files 113->135). Prior drift accumulated over multiple minor releases."
+    breaking_changes: []
+    deferred_to_follow_up:
+      - "AC2 — cross-lingual-conversion-copy.md KB (pt-BR <-> en-US adaptation patterns). Scoped out to ship ai-copy-human-loop-canon at canon quality rather than dilute two KBs."
+      - "AC3 — rsa-ads-copy-guide.md standalone stub (RSA character limits, pinning strategy). Lower priority, can ship as follow-up."
+
 quality_standards:
   reference_authors:
     - David Ogilvy (advertising fundamentals, Ogilvy on Advertising)