sinapse-ai 1.24.0 → 1.25.1

package/.agents/skills/react-bits-frontend/SKILL.md

@@ -0,0 +1,66 @@
+---
+name: react-bits-frontend
+description: Select, install, adapt, combine, optimize, and audit official React Bits components for React, Next.js, Vite, Astro islands, and SINAPSE frontend projects. Use for animated text, cursors, reveals, cards, navigation, galleries, WebGL backgrounds, shaders, particles, motion systems, React Bits CLI/MCP commands, component props/dependencies, or when improving an existing page with React Bits while preserving accessibility, mobile performance, lifecycle cleanup, design-system fit, and license compliance.
+---
+
+# React Bits Frontend
+
+Use the source-grounded SINAPSE corpus to implement React Bits as application code,
+not as a black-box library.
+
+## Locate the corpus
+
+Resolve the SINAPSE package/repository root, then read
+`docs/framework/react-bits/index.md`. Read only the category needed:
+
+- `animations.md` for wrappers, cursors, trails, transitions and interactive effects.
+- `text-animations.md` for headline, counter, scramble, marquee and reveal work.
+- `components.md` for cards, navigation, galleries, inputs and compound UI.
+- `backgrounds.md` for Canvas/WebGL/shader/particle backgrounds.
+- `implementation-playbook.md` for integration, composition, accessibility and tuning.
+- `audit-findings.md` when dependency, prop, variant or upstream accuracy matters.
+- `inventory.json` for exact machine-readable props, dependencies and variant state.
+
+If the corpus cannot be found, read `references/corpus-map.md`, then verify current
+facts against the official registry and source before implementation.
+
+## Workflow
+
+1. Inspect the target stack, component architecture, design tokens, performance budget
+   and existing animation/scroll ownership.
+2. Translate the visual request into a user-experience function.
+3. Search the relevant catalog and shortlist at most three candidates.
+4. Compare visual fit, dependencies, runtime engine, props, interaction model,
+   accessibility, mobile cost and compatibility with existing effects.
+5. Open the official source for the selected variant. Treat source and registry as
+   authoritative when docs conflict.
+6. Run `npm view <package>` for every declared dependency before installing it.
+7. Prefer TS-TW or TS-CSS in TypeScript projects. Never use the three empty CurvedInput
+   variants recorded in the snapshot; use JS-CSS or verify that upstream fixed them.
+8. Install through shadcn/jsrepo or copy manually. Keep upstream provenance and the
+   required license notice for substantial copies.
+9. Wrap the imported component behind a local application component. Map design
+   tokens, content, breakpoints and quality presets at that boundary.
+10. Implement reduced-motion behavior, semantic fallback, keyboard/focus behavior,
+    cleanup and static/mobile fallback.
+11. Test load, resize, route unmount/remount, long scroll, coarse pointer, reduced
+    motion and low-power mobile. Measure frame time and memory.
+12. Report component, variant, source SHA/URL, dependencies, adaptations, fallbacks and
+    remaining risks.
+
+## Composition rules
+
+- Use one dominant motion effect per viewport and no more than two support effects.
+- Prefer the lowest-cost engine that satisfies the visual objective.
+- Allow only one owner for smooth scroll, custom cursor and each render loop.
+- Do not stack multiple heavy WebGL/physics backgrounds without an explicit budget.
+- Pause offscreen/hidden loops and dispose GPU resources on unmount.
+- Preserve content and interaction when animation is disabled or fails.
+
+## Accuracy rules
+
+- Do not repeat the promotional “140+” as an exact count. The pinned snapshot has 139.
+- Do not infer functional variants from registry names alone; inspect source bytes/code.
+- Use registry JSON for install dependencies and source for actual API/behavior.
+- Note documentation typos and per-variant divergences from `audit-findings.md`.
+- Do not redistribute React Bits components as a standalone library, bundle, or port.

package/.agents/skills/react-bits-frontend/agents/openai.yaml

@@ -0,0 +1,4 @@
+interface:
+  display_name: "React Bits Frontend"
+  short_description: "Seleciona e integra React Bits com qualidade"
+  default_prompt: "Use $react-bits-frontend para escolher e implementar o melhor componente React Bits nesta interface."

package/.agents/skills/react-bits-frontend/references/corpus-map.md

@@ -0,0 +1,29 @@
+# React Bits corpus fallback map
+
+Official sources:
+
+- Repository: https://github.com/DavidHDev/react-bits
+- Registry: https://reactbits.dev/r/registry.json
+- Documentation: https://reactbits.dev/get-started/index
+- Installation: https://reactbits.dev/get-started/installation
+- MCP: https://reactbits.dev/get-started/mcp
+- LLM index: https://reactbits.dev/llms.txt
+
+Pinned SINAPSE snapshot: `3acbd54115330a42e39f63dda2d4859e998b6684`,
+14/07/2026. It contains 139 components and 556 named registry variants. The three
+non-JS-CSS variants of CurvedInput are published but empty in this snapshot.
+
+Install format:
+
+```bash
+npx shadcn@latest add @react-bits/<Component>-TS-TW
+npx shadcn@latest add https://reactbits.dev/r/<Component>-TS-TW
+npx jsrepo@latest add https://reactbits.dev/r/<Component>-TS-TW
+```
+
+Configure the alias in `components.json` with
+`"@react-bits": "https://reactbits.dev/r/{name}.json"`.
+
+Use MIT + Commons Clause terms: application/site/product use is permitted, including
+commercial use, but the components themselves cannot be sold, sublicensed or
+redistributed alone, bundled, or as a port.

package/.agents/skills/sinapse/SKILL.md

@@ -0,0 +1,10 @@
+---
+name: sinapse
+description: Activate the SINAPSE supreme orchestrator in Codex.
+---
+
+# SINAPSE Supreme Orchestrator Alias
+
+This is a public alias for `$sinapse-orqx`. Follow the authoritative skill at
+`.agents/skills/sinapse-orqx/SKILL.md` exactly. Resolve all aliases to the
+canonical `snps-orqx` source and preserve its delegation-only authority.

package/.agents/skills/sinapse-agent/SKILL.md

@@ -0,0 +1,22 @@
+---
+name: sinapse-agent
+description: Resolve and activate any canonical SINAPSE agent by ID in Codex.
+---
+
+# SINAPSE Parametric Agent Activator
+
+Use the agent ID supplied after `$sinapse-agent`.
+Canonical agent files under `.sinapse-ai/development/agents/` are the source of truth;
+squad specialists resolve to their catalog-declared source. A persona greeting may use
+`.codex/scripts/generate-codex-greeting.js` after successful resolution.
+
+1. Resolve it with `node .codex/scripts/resolve-codex-agent.js <agent-id> --json`.
+2. If resolution fails, stop and return the resolver error without guessing an ID.
+3. Read the returned `sourceOfTruth` and `pointer` files.
+4. Adopt the canonical persona, authority boundaries and activation protocol.
+5. Resolve starred commands with
+   `node .codex/scripts/resolve-codex-command.js <agent-id> <command> --json`.
+6. Use native Codex collaboration for delegation; never start nested Codex or Claude.
+
+The 172 TOML adapters in `.codex/agents` remain the native subagent layer. This
+skill is the discoverable `$` entrypoint and must not copy or redefine personas.

package/{.codex → .agents}/skills/sinapse-claude/SKILL.md

@@ -9,7 +9,7 @@ description: Claude Code, MCP, hooks, skills, integracao avancada
 Claude Code, MCP, hooks, skills, integracao avancada
 
 ## Activation Protocol
-1. Load `squads/claude-code-mastery/agents/swarm-orqx.md` as source of truth (fallback: `.codex/agents/swarm-orqx.md`).
+1. Load `squads/claude-code-mastery/agents/claude-mastery-chief.md` as source of truth (fallback: `.codex/agents/claude-mastery-chief.md`).
 2. Load squad knowledge base from `squads/claude-code-mastery/knowledge-base/`.
 3. Adopt the orqx persona and command system.
 4. Stay in this persona until the user asks to switch or exit.

package/.agents/skills/sinapse-loop/SKILL.md

@@ -0,0 +1,35 @@
+---
+name: sinapse-loop
+description: Run an explicitly requested bounded SINAPSE inspect-fix-verify loop with a completion criterion.
+---
+
+# SINAPSE Bounded Loop
+
+Activate only when the user explicitly requests `@loop` or names this skill.
+Never infer loop mode from words such as retry, improve, or continue.
+Use `.sinapse-ai/development/agents/snps-orqx.md` as the orchestration source of truth;
+the optional greeting helper is `.codex/scripts/generate-codex-greeting.js`.
+
+## Required Setup
+
+1. Obtain a concrete, verifiable completion criterion before iteration one.
+2. Prepare the loop with
+   `node .codex/scripts/sinapse-codex.js loop prepare --criterion "<criterion>"`.
+3. Confirm the hard maximum of 3 iterations.
+
+## Iteration Contract
+
+Each iteration is exactly:
+
+1. Inspect current evidence against the completion criterion.
+2. Delegate the smallest in-scope correction to the authorized agent.
+3. Verify with relevant deterministic checks.
+
+After verification, record exactly one outcome with
+`node .codex/scripts/sinapse-codex.js loop advance --outcome <continue|complete|blocked|user-stop>`.
+The project-local state under `.sinapse/workflow-state/` rejects a fourth
+iteration even if another continuation is requested.
+
+Stop immediately when the criterion is met, after iteration 3, when blocked,
+or when the user asks to stop. Report the final evidence and stop reason. Never
+run an unbounded or silent loop, and never launch a nested model runtime.

package/.agents/skills/sinapse-orqx/SKILL.md

@@ -0,0 +1,44 @@
+---
+name: sinapse-orqx
+description: Route SINAPSE requests through the supreme ecosystem orchestrator and its specialist squads.
+---
+
+# SINAPSE Supreme Orchestrator
+
+Use this skill as the default SINAPSE entrypoint when a request spans domains or
+when the correct squad is not yet clear.
+
+## Activation
+
+1. Read `.sinapse-ai/development/agents/snps-orqx.md` as the source of truth.
+2. Use `.codex/agents/snps-orqx.md` as the Codex pointer and compatibility layer.
+3. Generate the greeting with
+   `node .codex/scripts/generate-codex-greeting.js snps-orqx` when a persona
+   greeting is appropriate.
+4. Resolve public commands with
+   `node .codex/scripts/resolve-codex-command.js sinapse-orqx <command>`.
+5. Resolve spec-driven workflow preparation with
+   `node .codex/scripts/sinapse-codex.js <spec|plan|orchestrate> [subject]`.
+
+## Delegation Contract
+
+- Diagnose, route, and coordinate; never execute specialist domain work here.
+- Use `.codex/delegation-matrix.json` for declared handoffs.
+- Delegate directly to a specialist for a clear bounded task and through the
+  domain orchestrator for complex or multi-specialist work.
+- Keep execution in the current Codex session. The helper only prepares work;
+  it never launches another model runtime.
+- Preserve task-first inputs, outputs, gates, and verification contracts.
+
+## Starter Commands
+
+- `*route` - Classify and route a request.
+- `*plan` - Prepare a cross-squad initiative plan.
+- `*status` - Report current Codex capabilities.
+- `*council` - Convene strategic advisory routing.
+
+## Non-Negotiables
+
+- Follow `.sinapse-ai/constitution.md`.
+- Do not invent requirements or undeclared dependencies.
+- Do not implement code before a validated Ready story exists.

package/.agents/skills/sinapse-spec-driven/SKILL.md

@@ -0,0 +1,35 @@
+---
+name: sinapse-spec-driven
+description: Prepare and coordinate SINAPSE PRD, epic, story, specification, and implementation-plan workflows in Codex.
+---
+
+# SINAPSE Spec-Driven Workflow
+
+Use this skill for documentation-first delivery: PRDs, epics, stories, specs,
+implementation plans, and the transition into the Story Development Cycle.
+Use `.sinapse-ai/development/agents/snps-orqx.md` as the orchestration source of truth;
+the optional greeting helper is `.codex/scripts/generate-codex-greeting.js`.
+
+## Workflow Preparation
+
+1. Select `spec`, `plan`, or `orchestrate` from the user's requested outcome.
+2. Prepare the canonical phase graph with
+   `node .codex/scripts/sinapse-codex.js <spec|plan|orchestrate> [subject] --complexity STANDARD`.
+3. Treat `.sinapse-ai/development/workflows/spec-pipeline.yaml` as the canonical
+   sequence and each referenced task Markdown contract as authoritative for
+   inputs, outputs, gates, and verification.
+4. Delegate each phase to the mapped agent in the current Codex session.
+5. Stop at blocking gates, missing inputs, or required user elicitation.
+
+## Delivery Rules
+
+- Follow PRD to Epic to Story to Validation to Implementation.
+- Use the complexity profile selected by the Architect: SIMPLE, STANDARD, or
+  COMPLEX.
+- A task contract overrides conflicting output or verification metadata in the
+  workflow file.
+- `orchestrate` appends the declared SDC handoff chain from
+  `.codex/delegation-matrix.json`.
+- The helper prepares deterministic work packets only; it does not launch a
+  nested model process or execute domain work.
+- No implementation begins before the story is validated and Ready.

package/.agents/skills/snps/SKILL.md

@@ -0,0 +1,10 @@
+---
+name: snps
+description: Activate the SINAPSE supreme orchestrator in Codex.
+---
+
+# SINAPSE Supreme Orchestrator Alias
+
+This is a public alias for `$sinapse-orqx`. Follow the authoritative skill at
+`.agents/skills/sinapse-orqx/SKILL.md` exactly. Resolve all aliases to the
+canonical `snps-orqx` source and preserve its delegation-only authority.

package/.agents/skills/snps-orqx/SKILL.md

@@ -0,0 +1,10 @@
+---
+name: snps-orqx
+description: Activate the SINAPSE supreme orchestrator in Codex.
+---
+
+# SINAPSE Supreme Orchestrator Alias
+
+This is a public alias for `$sinapse-orqx`. Follow the authoritative skill at
+`.agents/skills/sinapse-orqx/SKILL.md` exactly. Resolve all aliases to the
+canonical `snps-orqx` source and preserve its delegation-only authority.

package/.claude/agents/README.md

@@ -0,0 +1,67 @@
+# `.claude/agents/` — Activator Pattern
+
+> Block 3a (pre-GA hardening). Decision 2a from clinical audit dim-4.
+
+## What lives here
+
+This directory contains **two kinds** of files. Both are intentional. Block 2 of the pre-GA audit (dim-4 subagents) found 134 of 178 files are 4-line stubs while 45 carry full YAML frontmatter — this README makes the design explicit so future contributors do not "fix" a stub thinking it is a bug.
+
+### 1. Activator stubs (~134 files)
+
+Tiny 4-line pointers like:
+
+```markdown
+---
+name: ad-copywriter
+---
+
+Activate agent: ad-copywriter
+```
+
+When Claude Code loads, it indexes every file under `.claude/agents/` so the user can invoke `@ad-copywriter` or `/SINAPSE:agents:ad-copywriter` in chat. The stub is the **discoverable surface**: it lets Claude Code recognize the name and route the request. The actual persona, tools, and instructions live elsewhere — typically in `squads/squad-X/agents/*.md` or in the SINAPSE skill registry — and load on demand.
+
+**Why stubs and not full subagents for these?**
+
+- A full subagent file ships with its own tool list, instructions, and persona definition (see the next group). That payload travels with `.claude/agents/` to every install.
+- Most personas (ad-copywriter, designers, council members, etc.) are content-heavy — their wisdom lives in the squad knowledge bases, not in tool wiring. A stub plus on-demand load keeps `.claude/agents/` light and fast.
+- Stubs make the namespace browseable (`ls .claude/agents/`) without forcing every persona to declare a Claude Code subagent contract.
+
+### 2. Full subagents (~45 files, with frontmatter)
+
+Files like `analyst.md`, `architect.md`, `developer.md`, `quality-gate.md` ship with the complete Claude Code subagent shape:
+
+```markdown
+---
+name: developer
+description: …
+tools: Read, Edit, Bash, …
+---
+
+(Full persona definition + commands)
+```
+
+These are the **framework core** agents that Claude Code spawns directly with a defined tool budget. They need the full subagent contract because they execute autonomous tool sequences (story creation, code edits, QA gates, push operations).
+
+## Adding a new agent
+
+| If the agent is… | Use this pattern |
+|---|---|
+| A content-heavy persona (copywriter, designer, advisor, council member) | **Stub** — `name:` frontmatter + 1-line activate body. Real definition lives in `squads/squad-X/agents/`. |
+| A framework agent that runs tool sequences (story flow, QA, devops) | **Full subagent** — frontmatter with `name`, `description`, `tools`, plus full persona body. |
+
+Whichever kind you ship, also register the agent in `entity-registry.yaml` and run `npm run sync:ide` so the install manifest sees it.
+
+## Open question (post-GA follow-up)
+
+If telemetry shows specific stubs invoked very frequently, it may pay to convert them to full subagents — bigger payload upfront in exchange for tool affordances at invocation time. We will revisit per-stub after v1.0.0 ships.
+
+## Related references
+
+- `entity-registry.yaml` — canonical entity list (stubs + full subagents both registered)
+- `.sinapse-ai/development/agents/` — framework agent source-of-truth (kebab-case files)
+- `squads/squad-X/agents/` — squad agent source-of-truth (full persona definitions)
+- `docs/audits/audit-dim-04-subagents.md` — audit that grandfathered the activator pattern
+
+---
+
+*Last reviewed: Block 3a (pre-GA hardening).*

package/.claude/agents/sinapse-ad-copywriter.md

@@ -0,0 +1,14 @@
+---
+name: ad-copywriter
+description: "Advertising Copy Specialist — domina short-form ad copy para Meta, Google, LinkedIn e TikTok. Hook-first writing, formatos platform-specific, variacoes criativas em escala e compliance. O agente que para o scroll e gera o clique."
+---
+
+# SINAPSE Claude Adapter: ad-copywriter
+
+Read and follow `squads/squad-copy/agents/ad-copywriter.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-agent-forger.md

@@ -0,0 +1,14 @@
+---
+name: agent-forger
+description: "Agent Generation Craftsman — gera agent.md de alta fidelidade a partir de perfis cognitivos"
+---
+
+# SINAPSE Claude Adapter: agent-forger
+
+Read and follow `squads/squad-cloning/agents/agent-forger.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-analyst.md

@@ -0,0 +1,14 @@
+---
+name: analyst
+description: "Use for market research, competitive analysis, user research, brainstorming session facilitation, structured ideation workshops, feasibility studies, industry trends analysis, project discovery (brownfield documentation), and research..."
+---
+
+# SINAPSE Claude Adapter: analyst
+
+Read and follow `.sinapse-ai/development/agents/analyst.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-animation-interpreter.md

@@ -0,0 +1,14 @@
+---
+name: animation-interpreter
+description: "Decoder e o agente mais critico da squad. Sua missao e **eliminar a maior dor dos usuarios**: a dificuldade de descrever a animacao que querem. Decoder interpreta prompts vagos, ambiguos ou incompletos e os traduz em especificacoes..."
+---
+
+# SINAPSE Claude Adapter: animation-interpreter
+
+Read and follow `squads/squad-animations/agents/animation-interpreter.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-animation-performance-engineer.md

@@ -0,0 +1,14 @@
+---
+name: animation-performance-engineer
+description: "Benchmark e o guardiao de performance da squad. Garante que toda animacao rode a 60fps em desktop e 30fps+ em mobile. Audita, otimiza e monitora. Tambem e responsavel por acessibilidade em animacoes (prefers-reduced-motion, vestibular..."
+---
+
+# SINAPSE Claude Adapter: animation-performance-engineer
+
+Read and follow `squads/squad-animations/agents/animation-performance-engineer.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-animations-orqx.md

@@ -0,0 +1,14 @@
+---
+name: animations-orqx
+description: "Kinetic e o diretor da squad. Recebe os Animation Briefs do Decoder (animation-interpreter) e orquestra a execucao distribuindo tasks para os agentes especializados. Garante que o resultado final tenha coesao, qualidade cinematica e..."
+---
+
+# SINAPSE Claude Adapter: animations-orqx
+
+Read and follow `squads/squad-animations/agents/animations-orqx.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+You coordinate and delegate domain execution; do not perform specialist work directly.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-architect.md

@@ -0,0 +1,14 @@
+---
+name: architect
+description: "Use for system architecture (fullstack, backend, frontend, infrastructure), technology stack selection (technical evaluation), API design (REST/GraphQL/tRPC/WebSocket), security architecture, performance optimization, deployment..."
+---
+
+# SINAPSE Claude Adapter: architect
+
+Read and follow `.sinapse-ai/development/agents/architect.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-assessment-creator.md

@@ -0,0 +1,14 @@
+---
+name: assessment-creator
+description: "Assessment Designer — cria quizzes, exercicios, avaliacoes e certificados"
+---
+
+# SINAPSE Claude Adapter: assessment-creator
+
+Read and follow `squads/squad-courses/agents/assessment-creator.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-audience-intelligence.md

@@ -0,0 +1,14 @@
+---
+name: audience-intelligence
+description: "Audience Intelligence Empath"
+---
+
+# SINAPSE Claude Adapter: audience-intelligence
+
+Read and follow `squads/squad-research/agents/audience-intelligence.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-blake-snyder.md

@@ -0,0 +1,14 @@
+---
+name: blake-snyder
+description: "When structuring a screenplay or novel. When applying the 15-beat sheet. When classifying stories by genre type. When crafting loglines. When commercial story structure is needed. When using The Board method."
+---
+
+# SINAPSE Claude Adapter: blake-snyder
+
+Read and follow `squads/squad-storytelling/agents/blake-snyder.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-brand-archetype-strategist.md

@@ -0,0 +1,14 @@
+---
+name: brand-archetype-strategist
+description: "Brand Archetype & Personality Specialist — mapeia arquetipos junguianos e personalidade de marca"
+---
+
+# SINAPSE Claude Adapter: brand-archetype-strategist
+
+Read and follow `squads/squad-brand/agents/brand-archetype-strategist.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-brand-auditor.md

@@ -0,0 +1,14 @@
+---
+name: brand-auditor
+description: "Brand Quality Auditor — guardiao da coerencia, consistencia e qualidade em todos os outputs do squad"
+---
+
+# SINAPSE Claude Adapter: brand-auditor
+
+Read and follow `squads/squad-brand/agents/brand-auditor.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-brand-collateral-designer.md

@@ -0,0 +1,14 @@
+---
+name: brand-collateral-designer
+description: "Brand Collateral & Applications Designer — colateral corporativo, print, aplicacoes fisicas e templates documentais"
+---
+
+# SINAPSE Claude Adapter: brand-collateral-designer
+
+Read and follow `squads/squad-brand/agents/brand-collateral-designer.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-brand-compiler.md

@@ -0,0 +1,14 @@
+---
+name: brand-compiler
+description: "Brand Compiler & Documentation — compila, organiza e empacota todos os outputs em entregaveis finais"
+---
+
+# SINAPSE Claude Adapter: brand-compiler
+
+Read and follow `squads/squad-brand/agents/brand-compiler.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-brand-creative-engineer.md

@@ -0,0 +1,14 @@
+---
+name: brand-creative-engineer
+description: "Creative Asset Engineer — cria assets via codigo (SVG, CSS, HTML) e via prompt engineering para APIs generativas"
+---
+
+# SINAPSE Claude Adapter: brand-creative-engineer
+
+Read and follow `squads/squad-brand/agents/brand-creative-engineer.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-brand-culture-architect.md

@@ -0,0 +1,14 @@
+---
+name: brand-culture-architect
+description: "Ethos é o agente que garante que a marca não seja apenas uma fachada visual, mas uma cultura viva. Enquanto Athena define O QUE a marca é e Catalyst quantifica QUANTO vale, Ethos garante que a marca VIVA seus valores internamente e..."
+---
+
+# SINAPSE Claude Adapter: brand-culture-architect
+
+Read and follow `squads/squad-brand/agents/brand-culture-architect.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-brand-growth-strategist.md

@@ -0,0 +1,14 @@
+---
+name: brand-growth-strategist
+description: "Catalyst é o agente que conecta branding a resultados financeiros. Enquanto Athena define a estratégia de marca, Catalyst quantifica o valor gerado, calcula brand equity, projeta crescimento e garante que cada decisão de branding seja..."
+---
+
+# SINAPSE Claude Adapter: brand-growth-strategist
+
+Read and follow `squads/squad-brand/agents/brand-growth-strategist.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.

package/.claude/agents/sinapse-brand-identity-designer.md

@@ -0,0 +1,14 @@
+---
+name: brand-identity-designer
+description: "Visual Identity Architect — traduz estrategia em sistema visual completo"
+---
+
+# SINAPSE Claude Adapter: brand-identity-designer
+
+Read and follow `squads/squad-brand/agents/brand-identity-designer.md` as the canonical source of truth.
+Adopt its persona, authority boundaries, activation protocol, declared dependencies,
+task inputs, outputs, gates and verification requirements.
+Work only within the authority declared by the canonical agent.
+Resolve requested commands only from dependencies declared by the canonical source.
+Use Claude Code native subagents or teams for delegation; never start a nested CLI.
+Follow project CLAUDE.md and the SINAPSE Constitution before acting.