heimdall-agent 1.1.0 → 1.2.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (41) hide show
  1. package/README.md +13 -13
  2. package/README.pt-BR.md +36 -22
  3. package/agents/code-sentinel.md +7 -1
  4. package/agents/modernizer.md +4 -2
  5. package/agents/orchestrator.md +20 -7
  6. package/agents/reviewer.md +2 -0
  7. package/config/agents.json +11 -1
  8. package/config/mcp.json +1 -1
  9. package/config/profiles.json +1 -1
  10. package/config/skills.json +762 -683
  11. package/docs/SKILLS.md +96 -74
  12. package/docs/runs/skill-audit/2026-11-12-compare-report.json +1028 -0
  13. package/docs/runs/skill-audit/2026-11-12-upstream-audit.json +275 -0
  14. package/doctrine/AGENTS.md +4 -1
  15. package/lib/catalog-sync.cjs +8 -3
  16. package/package.json +1 -1
  17. package/playbooks/01-interactive-grilling-gate.md +14 -10
  18. package/playbooks/03-adversarial-code-review.md +12 -4
  19. package/playbooks/05-canonical-plan-template.md +3 -1
  20. package/setup.mjs +2 -2
  21. package/skills/antislop-copywriting/SKILL.md +36 -1
  22. package/skills/brainstorming/SKILL.md +60 -38
  23. package/skills/codebase-design/DEEPENING.md +37 -0
  24. package/skills/codebase-design/DESIGN-IT-TWICE.md +44 -0
  25. package/skills/codebase-design/SKILL.md +114 -0
  26. package/skills/domain-modeling/ADR-FORMAT.md +47 -0
  27. package/skills/domain-modeling/GLOSSARY-FORMAT.md +60 -0
  28. package/skills/domain-modeling/SKILL.md +74 -0
  29. package/skills/executing-plans/SKILL.md +6 -4
  30. package/skills/grill-me/SKILL.md +65 -0
  31. package/skills/handoff/SKILL.md +64 -0
  32. package/skills/improve-codebase-architecture/HTML-REPORT.md +123 -0
  33. package/skills/improve-codebase-architecture/SKILL.md +71 -0
  34. package/skills/owasp-security/reference/config-and-supply-chain.md +346 -344
  35. package/skills/owasp-security/reference/languages.md +378 -372
  36. package/skills/owasp-security/reference/owasp-report.md +1144 -1140
  37. package/skills/owasp-security/reference/review-checklist.md +134 -132
  38. package/skills/subagent-driven-development/SKILL.md +7 -4
  39. package/skills/to-spec/SKILL.md +75 -0
  40. package/skills/to-tickets/SKILL.md +105 -0
  41. package/skills/wayfinder/SKILL.md +128 -0
package/README.md CHANGED
@@ -4,13 +4,13 @@
4
4
  <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-blue.svg?style=flat-square" alt="License: MIT" /></a>
5
5
  <a href="https://nodejs.org"><img src="https://img.shields.io/badge/node-%3E%3D18.0.0-339933.svg?style=flat-square&logo=node.js&logoColor=white" alt="Node >= 18" /></a>
6
6
  <a href="#10-predefined-agent-personas"><img src="https://img.shields.io/badge/agents-10%20predefined-orange.svg?style=flat-square" alt="Agents: 10 Predefined" /></a>
7
- <a href="#skill-inventory-56-modules-across-24-toolkits"><img src="https://img.shields.io/badge/skills-24%20toolkits%20%2856%20modules%29-blueviolet.svg?style=flat-square" alt="Skills: 24 Toolkits (56 Modules)" /></a>
7
+ <a href="#skill-inventory-64-modules-across-25-toolkits"><img src="https://img.shields.io/badge/skills-25%20toolkits%20%2864%20modules%29-blueviolet.svg?style=flat-square" alt="Skills: 25 Toolkits (64 Modules)" /></a>
8
8
  <a href="#supported-tools--runtimes"><img src="https://img.shields.io/badge/runtimes-Antigravity%20%7C%20Claude%20%7C%20Codex%20%7C%20Hermes%20%7C%20OpenClaw-black.svg?style=flat-square" alt="Runtimes" /></a>
9
9
  </p>
10
10
 
11
11
  🌐 **Language / Idioma:** **English** | [Português (Brasil)](README.pt-BR.md)
12
12
 
13
- Heimdall is a portable bootstrap toolkit and quality sentinel for multi-agent software engineering. It enforces planning gates before code generation, prevents AI slop, isolates module boundaries, and synchronizes **10 specialized agent personas** and **56 modular skills** across **Google Antigravity**, **Claude Code**, **OpenAI Codex CLI**, **Hermes Agent**, **OpenClaw**, **OpenDesign**, and **Claude Desktop**.
13
+ Heimdall is a portable bootstrap toolkit and quality sentinel for multi-agent software engineering. It enforces planning gates before code generation, prevents AI slop, isolates module boundaries, and synchronizes **10 specialized agent personas** and **64 modular skills** across **Google Antigravity**, **Claude Code**, **OpenAI Codex CLI**, **Hermes Agent**, **OpenClaw**, **OpenDesign**, and **Claude Desktop**.
14
14
 
15
15
  ---
16
16
 
@@ -34,7 +34,7 @@ Heimdall is a portable bootstrap toolkit and quality sentinel for multi-agent so
34
34
  **Heimdall** is an environment bootstrapper and multi-agent coordination layer for AI coding tools.
35
35
 
36
36
  Instead of manually maintaining scattered custom instructions, system prompts, and configuration files across different AI clients, Heimdall provides a single source of truth:
37
- * **Modular Skills**: 56 task-activated capabilities (such as database design, security audits, design token enforcement, progressive disclosure, and debugging) that load on demand instead of wasting prompt tokens.
37
+ * **Modular Skills**: 64 task-activated capabilities (such as database design, security audits, design token enforcement, progressive disclosure, and debugging) that load on demand instead of wasting prompt tokens.
38
38
  * **Specialized Agent Personas**: 10 predefined roles with isolated context windows (`orchestrator`, `tester`, `frontend-designer`, `mobile-designer`, `reviewer`, `security-auditor`, `database-architect`, `code-sentinel`, `modernizer`, `cloud-engineer`).
39
39
  * **Universal Doctrine**: Central rules (`AGENTS.md`) synchronized across all harnesses, enforcing planning gates, test-driven development (TDD), and minimal surgical code edits.
40
40
 
@@ -82,7 +82,7 @@ When Heimdall bootstraps, it establishes live junctions (Windows) or symlinks (P
82
82
 
83
83
  | Source in Heimdall | Deployed Location | Purpose |
84
84
  | :--- | :--- | :--- |
85
- | `skills/` | `~/.<tool>/skills/` | 56 domain toolkits loaded progressively on demand. |
85
+ | `skills/` | `~/.<tool>/skills/` | 64 domain toolkits loaded progressively on demand. |
86
86
  | `agents/` | `~/.<tool>/agents/` | 10 specialized agent persona definitions. |
87
87
  | `hooks/` | `~/.claude/hooks/` | Event-driven lifecycle hooks for Claude Code CLI. |
88
88
  | `hooks/git/` | `.git/hooks/` (optional) | Universal native Git pre-push hook for all tools. |
@@ -152,7 +152,7 @@ Select a profile tier based on your machine and workflow needs:
152
152
 
153
153
  | Profile | Skills | Predefined Agents | Description | Best For |
154
154
  | :--- | :---: | :---: | :--- | :--- |
155
- | **`full`** *(Default)* | 56 | 10 | Complete engineering suite: planning, design, security, database, hygiene. | Everyday development across all tools |
155
+ | **`full`** *(Default)* | 64 | 10 | Complete engineering suite: planning, design, security, database, hygiene. | Everyday development across all tools |
156
156
  | **`minimal`** | 16 | 4 | Lean core: `orchestrator`, `tester`, `code-sentinel`, `reviewer`. | Lightweight laptops, smaller context windows |
157
157
  | **`security`** | 12 | 4 | AppSec suite: `orchestrator`, `security-auditor`, `tester`, `reviewer`. | Penetration testing & pre-deployment audits |
158
158
  | **`design`** | 12 | 4 | Frontend & mobile: `orchestrator`, `frontend-designer`, `mobile-designer`, `code-sentinel`. | UI/UX design and design-system work |
@@ -289,25 +289,25 @@ All personas declare `model: inherit` to run on whichever active model or vendor
289
289
 
290
290
  | Persona | Capability Tier | Role & Responsibilities | Bound Skills |
291
291
  | :--- | :--- | :--- | :--- |
292
- | **`orchestrator`** | `deep_reasoning` | Planning gatekeeper (`/grill-me`), requirements interrogation, modular task decomposition, Playwright E2E verification enforcement, and subagent coordination. | `brainstorming`, `writing-plans`, `pstack-discipline`, `subagent-driven-development`, `dispatching-parallel-agents`, `using-git-worktrees`, `writing-skills`, `executing-plans`, `herdr` |
292
+ | **`orchestrator`** | `deep_reasoning` | Planning gatekeeper (`/grill-me`), requirements interrogation, modular task decomposition, Playwright E2E verification enforcement, and subagent coordination. | `grill-me`, `brainstorming`, `writing-plans`, `to-spec`, `to-tickets`, `wayfinder`, `domain-modeling`, `handoff`, `pstack-discipline`, `subagent-driven-development`, `dispatching-parallel-agents`, `using-git-worktrees`, `writing-skills`, `executing-plans`, `herdr` |
293
293
  | **`frontend-designer`** | `high_polish_vision` | Enforces design read, typography hierarchy, token mappings, and anti-slop UI aesthetics for web surfaces. | `taste-skill`, `brandkit`, `minimalist-skill`, `brutalist-skill`, `soft-skill`, `antislop-ui`, `antislop-human`, `stitch-skill`, `image-to-code-skill`, `imagegen-frontend-web`, `redesign-skill` |
294
294
  | **`mobile-designer`** | `responsive_flex_safety` | Mobile viewport safety (320px–430px), 48px touch targets, zero-clipping overflow flex containers, and responsive assets. | `antislop-layoutmobile`, `antislop-human`, `antislop-ui`, `imagegen-frontend-mobile`, `taste-skill` |
295
295
  | **`tester`** | `fast_verification` | Enforces strict Test-Driven Development (TDD) and browser verification (Playwright). Rejects pseudo-verification (`curl`, `grep`, `tsc`). | `test-driven-development`, `verification-before-completion`, `systematic-debugging`, `executing-plans` |
296
296
  | **`security-auditor`** | `deep_assurance_audit` | Dynamic differential HTTP probing across auth levels, OWASP Top 10 defenses, and Trail of Bits insecure defaults detection. | `strix-security`, `security-audit-sweep`, `owasp-security`, `cloudflare-security-audit`, `snyk-agent-scan`, `trailofbits-audit-context`, `trailofbits-insecure-defaults`, `trailofbits-triage-brocards` |
297
297
  | **`database-architect`** | `schema_migration_safety` | Governs non-destructive migrations, normalization, query plan audits (`EXPLAIN ANALYZE`), and connection pooling (Postgres/Supabase, SQLite, MySQL). | `postgres-best-practices`, `systematic-debugging`, `karpathy-guidelines` |
298
- | **`code-sentinel`** | `surgical_restraint` | Karpathy simplicity: writes minimal surgical code, strips boilerplate comments, purges defensive dead code, and enforces anti-slop standards. | `karpathy-guidelines`, `clean-code`, `ponytail`, `antislop`, `antislop-code`, `antislop-copywriting`, `antislop-human`, `finishing-a-development-branch` |
299
- | **`reviewer`** | `adversarial_diff_eval` | Performs adversarial diff analysis, catches silent regressions and swallowed errors, executes thermo maintainability reviews, and coordinates cross-model validation reviews. | `communication-doctrine`, `thermo`, `requesting-code-review`, `receiving-code-review`, `claude-collaborator`, `gemini-collaborator`, `codex-collaborator`, `using-superpowers`, `diagnosing-superpowers`, `output-skill` |
300
- | **`modernizer`** | `legacy_migration_safety` | Executes legacy codebase transformations, runtime upgrades, and Strangler Fig migrations guarded by Golden Master characterization tests. | `test-driven-development`, `verification-before-completion`, `systematic-debugging`, `karpathy-guidelines`, `project-hygiene`, `output-skill` |
298
+ | **`code-sentinel`** | `surgical_restraint` | Karpathy simplicity: writes minimal surgical code, strips boilerplate comments, purges defensive dead code, enforces deep modules, and applies anti-slop standards. | `karpathy-guidelines`, `clean-code`, `codebase-design`, `ponytail`, `antislop`, `antislop-code`, `antislop-copywriting`, `antislop-human`, `finishing-a-development-branch` |
299
+ | **`reviewer`** | `adversarial_diff_eval` | Performs adversarial diff analysis, audits module depth and leaky abstractions, catches silent regressions, executes thermo maintainability reviews, and coordinates cross-model validation reviews. | `communication-doctrine`, `thermo`, `codebase-design`, `requesting-code-review`, `receiving-code-review`, `claude-collaborator`, `gemini-collaborator`, `codex-collaborator`, `using-superpowers`, `diagnosing-superpowers`, `output-skill` |
300
+ | **`modernizer`** | `legacy_migration_safety` | Executes legacy codebase transformations, runtime upgrades, dependency uncoupling, and Strangler Fig migrations guarded by Golden Master characterization tests. | `test-driven-development`, `verification-before-completion`, `systematic-debugging`, `improve-codebase-architecture`, `codebase-design`, `karpathy-guidelines`, `project-hygiene`, `output-skill` |
301
301
  | **`cloud-engineer`** | `infra_container_safety` | Governs containerization (Docker, Compose), Kubernetes manifests, Infrastructure as Code (Terraform), reverse proxies (Caddy, Nginx), CI/CD automation, and terminating verification. | `communication-doctrine`, `karpathy-guidelines`, `clean-code`, `systematic-debugging`, `project-hygiene`, `progressive-disclosure` |
302
302
 
303
303
  ---
304
304
 
305
- <a id="skill-inventory-56-modules-across-24-toolkits"></a>
306
- ### Skill Inventory (56 Modules Across 24 Toolkits)
305
+ <a id="skill-inventory-64-modules-across-25-toolkits"></a>
306
+ ### Skill Inventory (64 Modules Across 25 Toolkits)
307
307
 
308
- Heimdall organizes capabilities into **5 Core Multi-Skill Suites (37 modules)** and **19 Standalone Toolkits (19 modules)**:
308
+ Heimdall organizes capabilities into **6 Core Multi-Skill Suites (45 modules)** and **19 Standalone Toolkits (19 modules)**:
309
309
 
310
- #### Core Multi-Skill Suites (5 Suites / 37 Modules)
310
+ #### Core Multi-Skill Suites (6 Suites / 45 Modules)
311
311
 
312
312
  <details open>
313
313
  <summary><b>1. Superpowers Engineering Suite (15 modules) — Jesse Vincent / obra</b></summary>
package/README.pt-BR.md CHANGED
@@ -4,13 +4,13 @@
4
4
  <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-blue.svg?style=flat-square" alt="License: MIT" /></a>
5
5
  <a href="https://nodejs.org"><img src="https://img.shields.io/badge/node-%3E%3D18.0.0-339933.svg?style=flat-square&logo=node.js&logoColor=white" alt="Node >= 18" /></a>
6
6
  <a href="#10-personas-de-agentes-predefinidas"><img src="https://img.shields.io/badge/agents-10%20predefinidos-orange.svg?style=flat-square" alt="Agentes: 10 Predefinidos" /></a>
7
- <a href="#inventario-de-skills-56-modulos-em-24-toolkits"><img src="https://img.shields.io/badge/skills-24%20toolkits%20%2856%20m%C3%B3dulos%29-blueviolet.svg?style=flat-square" alt="Skills: 24 Toolkits (56 Módulos)" /></a>
7
+ <a href="#inventario-de-skills-64-modulos-em-25-toolkits"><img src="https://img.shields.io/badge/skills-25%20toolkits%20%2864%20m%C3%B3dulos%29-blueviolet.svg?style=flat-square" alt="Skills: 25 Toolkits (64 Módulos)" /></a>
8
8
  <a href="#ferramentas-e-runtimes-suportados"><img src="https://img.shields.io/badge/runtimes-Antigravity%20%7C%20Claude%20%7C%20Codex%20%7C%20Hermes%20%7C%20OpenClaw-black.svg?style=flat-square" alt="Runtimes" /></a>
9
9
  </p>
10
10
 
11
11
  🌐 **Idioma / Language:** [English](README.md) | **Português (Brasil)**
12
12
 
13
- O Heimdall é um toolkit portátil de bootstrap e sentinela de qualidade para engenharia de software multi-agente. Ele impõe portões de planejamento antes da geração de código, previne *AI slop* (código e texto genérico de IA), isola fronteiras de módulos e sincroniza **10 personas de agentes especializados** e **56 skills modulares** entre **Google Antigravity**, **Claude Code**, **OpenAI Codex CLI**, **Hermes Agent**, **OpenClaw**, **OpenDesign** e **Claude Desktop**.
13
+ O Heimdall é um toolkit portátil de bootstrap e sentinela de qualidade para engenharia de software multi-agente. Ele impõe portões de planejamento antes da geração de código, previne *AI slop* (código e texto genérico de IA), isola fronteiras de módulos e sincroniza **10 personas de agentes especializados** e **64 skills modulares** entre **Google Antigravity**, **Claude Code**, **OpenAI Codex CLI**, **Hermes Agent**, **OpenClaw**, **OpenDesign** e **Claude Desktop**.
14
14
 
15
15
  ---
16
16
 
@@ -34,7 +34,7 @@ O Heimdall é um toolkit portátil de bootstrap e sentinela de qualidade para en
34
34
  O **Heimdall** é um bootstrapper de ambiente e camada de coordenação multi-agente para ferramentas de programação com IA.
35
35
 
36
36
  Em vez de manter manualmente instruções customizadas, prompts de sistema e arquivos de configuração espalhados em diferentes clientes de IA, o Heimdall fornece uma fonte única da verdade (*Single Source of Truth*):
37
- * **Skills Modulares**: 56 capacidades ativadas sob demanda por tarefa (como modelagem de banco de dados, auditoria de segurança, aplicação de design tokens, divulgação progressiva e depuração sistemática) que carregam apenas quando necessário, economizando tokens de contexto.
37
+ * **Skills Modulares**: 64 capacidades ativadas sob demanda por tarefa (como modelagem de banco de dados, auditoria de segurança, aplicação de design tokens, divulgação progressiva e depuração sistemática) que carregam apenas quando necessário, economizando tokens de contexto.
38
38
  * **Personas de Agentes Especializados**: 10 papéis predefinidos com janelas de contexto isoladas (`orchestrator`, `tester`, `frontend-designer`, `mobile-designer`, `reviewer`, `security-auditor`, `database-architect`, `code-sentinel`, `modernizer`, `cloud-engineer`).
39
39
  * **Doutrina Universal**: Regras centrais (`AGENTS.md`) sincronizadas em todas as ferramentas, exigindo portões de planejamento, desenvolvimento orientado a testes (TDD) e edições cirúrgicas mínimas.
40
40
 
@@ -82,28 +82,28 @@ Durante o bootstrap, o Heimdall cria *junctions* (Windows) ou *symlinks* (POSIX)
82
82
 
83
83
  | Origem no Heimdall | Local Implantado | Finalidade |
84
84
  | :--- | :--- | :--- |
85
- | `skills/` | `~/.<ferramenta>/skills/` | 56 toolkits de domínio carregados progressivamente sob demanda. |
85
+ | `skills/` | `~/.<ferramenta>/skills/` | 64 toolkits de domínio carregados progressivamente sob demanda. |
86
86
  | `agents/` | `~/.<ferramenta>/agents/` | 10 definições de personas de agentes especializados. |
87
87
  | `hooks/` | `~/.claude/hooks/` | Hooks de ciclo de vida orientados a eventos para o Claude Code CLI. |
88
88
  | `hooks/git/` | `.git/hooks/` (opcional) | Hook Git `pre-push` nativo universal para todas as ferramentas. |
89
89
  | `doctrine/AGENTS.md` | `~/.claude/CLAUDE.md`<br>`~/.gemini/GEMINI.md`<br>`~/.codex/AGENTS.md`<br>`~/.hermes/AGENTS.md` | Doutrina universal entre agentes e restrições operacionais. |
90
90
  | `playbooks/` | `~/.<ferramenta>/playbooks/` | Playbooks operacionais (portão de perguntas, TDD, revisões, planos, backlog & fluxo CR). |
91
91
  | `commands/` | `~/.claude/commands/` | Comandos slash (`/orchestrator`, `/orchestrate`, `/plan`, `/grill-me`). |
92
- | `tools/` | Diretório no `PATH` | Scripts de ponte entre modelos (`claude-bridge`, `gemini-bridge`, `codex-bridge`) e CLI `heimdall`. |
92
+ | `tools/` | Diretório no `PATH` | Executável da CLI `heimdall` e utilitários de inicialização de ambiente. |
93
93
  | `docs/designs/` | Raiz do workspace | Localização canônica de planos (`YYYY-MM-DD <nome>.md`); planos verificados vão para `old/`. |
94
94
  | `docs/backlog.md` | Raiz do workspace | Ledger canônico de backlog do produto e Change Requests. |
95
95
 
96
- ### 2. O Modelo Elefante-Peixe-Dourado do Google (EGM)
96
+ ### 2. Separação de Contexto: Padrão Elephant-Goldfish (Google EGM)
97
97
 
98
98
  O Heimdall adota a divisão central de trabalho do [`google/elephant-goldfish-model`](https://github.com/google/elephant-goldfish-model):
99
99
 
100
- * **O Elefante (Coordenador)**: A sessão principal e a persona [`orchestrator`](#10-personas-de-agentes-predefinidas) retêm a memória de longo prazo, a narrativa do projeto e o histórico profundo de contexto.
101
- * **O Peixe-Dourado (Revisores e Testadores)**: Subagentes despachados para revisão de código (`reviewer`) e diagnóstico de bugs (`tester`) são intencionalmente isolados do histórico de conversa e dos debates do autor. Eles avaliam `git diffs` brutos e critérios de aceite diretamente, prevenindo viés de confirmação.
102
- * **Escada de Fronteira em 3 Níveis (3-Tier Boundary Ladder)**:
103
- - **Tier 1 (Spike / Bugfix < 50 LOC)**: Zero cerimônia. Código cirúrgico mínimo com teste falhando primeiro.
104
- - **Tier 2 (Feature Delimitada)**: Verificação rápida de design no chat com revisões de crítica e prontidão.
105
- - **Tier 3 (Arquitetural)**: Especificação gravada em disco (`docs/designs/<feature>.md`), revisão em 4 estágios (Zoom Out, Compreensão, Crítica, Prontidão) e ledger de convergência de até 5 rodadas.
106
- * **Execução Multi-Modelo Adaptativa**: No Herdr (`HERDR_ENV=1`), o Heimdall roteia revisões entre modelos distintos (ex.: Elefante Gemini validado por Claude/Codex) via multiplexação nativa do Herdr; caso contrário, recorre automaticamente a subagentes independentes no mesmo modelo.
100
+ * **O Elefante (Orquestrador / Memória Longa)**: A sessão principal e a persona [`orchestrator`](#10-personas-de-agentes-predefinidas) preservam a memória contínua, as decisões consolidadas da arquitetura e o contexto de longo prazo.
101
+ * **O Goldfish (Revisores e Testadores com Contexto Cego)**: Subagentes acionados para revisão (`reviewer`) e testes (`tester`) recebem exclusivamente o `git diff` e os critérios de aceitação, sem acesso a justificativas ou debates anteriores. Essa amnésia intencional elimina o viés de confirmação e elogios vazios.
102
+ * **Governança Gradual em 3 Níveis**:
103
+ - **Nível 1 (Correções Rápidas / Bugfix < 50 linhas)**: Sem burocracia. Edição cirúrgica mínima orientada por testes automatizados.
104
+ - **Nível 2 (Funcionalidades Delimitadas)**: Alinhamento ágil de arquitetura com verificação rápida de viabilidade e prontidão.
105
+ - **Nível 3 (Arquitetura e Alto Impacto)**: Especificação formal persistida em disco (`docs/designs/<feature>.md`), revisão em 4 etapas (Visão Geral, Compreensão, Crítica e Prontidão) e ciclo de convergência com até 5 rodadas.
106
+ * **Execução Multi-Modelo Adaptativa**: No ambiente Herdr (`HERDR_ENV=1`), o Heimdall distribui revisões entre modelos distintos (por exemplo, código gerado pelo Gemini sendo auditado pelo Claude ou Codex) em subpainéis nativos; fora do Herdr, recorre a subagentes isolados no mesmo modelo.
107
107
 
108
108
  ### 3. Hooks de Ciclo de Vida Orientados a Eventos
109
109
 
@@ -152,7 +152,7 @@ Escolha o perfil de acordo com sua máquina e fluxo de trabalho:
152
152
 
153
153
  | Perfil | Skills | Agentes Predefinidos | Descrição | Ideal Para |
154
154
  | :--- | :---: | :---: | :--- | :--- |
155
- | **`full`** *(Padrão)* | 56 | 10 | Suíte completa de engenharia: planejamento, design, segurança, banco de dados e higiene. | Desenvolvimento diário em todas as ferramentas |
155
+ | **`full`** *(Padrão)* | 64 | 10 | Suíte completa de engenharia: planejamento, design, segurança, banco de dados e higiene. | Desenvolvimento diário em todas as ferramentas |
156
156
  | **`minimal`** | 16 | 4 | Núcleo enxuto: `orchestrator`, `tester`, `code-sentinel`, `reviewer`. | Notebooks leves, janelas de contexto menores |
157
157
  | **`security`** | 12 | 4 | Suíte AppSec: `orchestrator`, `security-auditor`, `tester`, `reviewer`. | Pentest e auditorias pré-deploy |
158
158
  | **`design`** | 12 | 4 | Frontend e mobile: `orchestrator`, `frontend-designer`, `mobile-designer`, `code-sentinel`. | Design de UI/UX e design systems |
@@ -288,25 +288,25 @@ Todas as personas declaram `model: inherit` para rodar em qualquer modelo ou pro
288
288
 
289
289
  | Persona | Nível de Capacidade | Papel e Responsabilidades | Skills Vinculadas |
290
290
  | :--- | :--- | :--- | :--- |
291
- | **`orchestrator`** | `deep_reasoning` | Guardião do planejamento (`/grill-me`), questionamento de requisitos, decomposição modular de tarefas, exigência de verificação Playwright E2E e coordenação de subagentes. | `brainstorming`, `writing-plans`, `pstack-discipline`, `subagent-driven-development`, `dispatching-parallel-agents`, `using-git-worktrees`, `writing-skills`, `executing-plans`, `herdr` |
291
+ | **`orchestrator`** | `deep_reasoning` | Guardião do planejamento (`/grill-me`), questionamento de requisitos, decomposição modular de tarefas, exigência de verificação Playwright E2E e coordenação de subagentes. | `grill-me`, `brainstorming`, `writing-plans`, `to-spec`, `to-tickets`, `wayfinder`, `domain-modeling`, `handoff`, `pstack-discipline`, `subagent-driven-development`, `dispatching-parallel-agents`, `using-git-worktrees`, `writing-skills`, `executing-plans`, `herdr` |
292
292
  | **`frontend-designer`** | `high_polish_vision` | Define leitura de design, hierarquia tipográfica, mapeamento de tokens e estética de UI anti-slop para web. | `taste-skill`, `brandkit`, `minimalist-skill`, `brutalist-skill`, `soft-skill`, `antislop-ui`, `antislop-human`, `stitch-skill`, `image-to-code-skill`, `imagegen-frontend-web`, `redesign-skill` |
293
293
  | **`mobile-designer`** | `responsive_flex_safety` | Segurança de viewport mobile (320px–430px), alvos de toque de 48px, containers flex sem corte de overflow e assets responsivos. | `antislop-layoutmobile`, `antislop-human`, `antislop-ui`, `imagegen-frontend-mobile`, `taste-skill` |
294
294
  | **`tester`** | `fast_verification` | Impõe TDD estrito e verificação real em navegador (Playwright). Rejeita pseudo-verificação (`curl`, `grep`, `tsc`). | `test-driven-development`, `verification-before-completion`, `systematic-debugging`, `executing-plans` |
295
295
  | **`security-auditor`** | `deep_assurance_audit` | Sondagem HTTP diferencial dinâmica entre níveis de autenticação, defesas OWASP Top 10 e detecção de padrões inseguros da Trail of Bits. | `strix-security`, `security-audit-sweep`, `owasp-security`, `cloudflare-security-audit`, `snyk-agent-scan`, `trailofbits-audit-context`, `trailofbits-insecure-defaults`, `trailofbits-triage-brocards` |
296
296
  | **`database-architect`** | `schema_migration_safety` | Governa migrações não destrutivas, normalização, auditoria de planos de query (`EXPLAIN ANALYZE`) e pool de conexões (Postgres/Supabase, SQLite, MySQL). | `postgres-best-practices`, `systematic-debugging`, `karpathy-guidelines` |
297
- | **`code-sentinel`** | `surgical_restraint` | Simplicidade Karpathy: escreve código cirúrgico mínimo, remove comentários redundantes, elimina código morto defensivo e aplica padrões anti-slop. | `karpathy-guidelines`, `clean-code`, `ponytail`, `antislop`, `antislop-code`, `antislop-copywriting`, `antislop-human`, `finishing-a-development-branch` |
298
- | **`reviewer`** | `adversarial_diff_eval` | Realiza análise adversarial de diffs, captura regressões silenciosas e erros engolidos, executa auditorias thermo de manutenibilidade e coordena revisões cruzadas entre modelos. | `communication-doctrine`, `thermo`, `requesting-code-review`, `receiving-code-review`, `claude-collaborator`, `gemini-collaborator`, `codex-collaborator`, `using-superpowers`, `diagnosing-superpowers`, `output-skill` |
299
- | **`modernizer`** | `legacy_migration_safety` | Executa transformações de código legado, upgrades de runtime e migrações Strangler Fig protegidas por testes de caracterização Golden Master. | `test-driven-development`, `verification-before-completion`, `systematic-debugging`, `karpathy-guidelines`, `project-hygiene`, `output-skill` |
297
+ | **`code-sentinel`** | `surgical_restraint` | Simplicidade Karpathy: escreve código cirúrgico mínimo, remove comentários redundantes, elimina código morto defensivo, exige módulos profundos e aplica padrões anti-slop. | `karpathy-guidelines`, `clean-code`, `codebase-design`, `ponytail`, `antislop`, `antislop-code`, `antislop-copywriting`, `antislop-human`, `finishing-a-development-branch` |
298
+ | **`reviewer`** | `adversarial_diff_eval` | Realiza análise adversarial de diffs, audita profundidade de módulos e abstrações rasas, captura regressões silenciosas, executa auditorias thermo e coordena revisões cruzadas entre modelos. | `communication-doctrine`, `thermo`, `codebase-design`, `requesting-code-review`, `receiving-code-review`, `claude-collaborator`, `gemini-collaborator`, `codex-collaborator`, `using-superpowers`, `diagnosing-superpowers`, `output-skill` |
299
+ | **`modernizer`** | `legacy_migration_safety` | Executa transformações de código legado, upgrades de runtime, desacoplamento de dependências e migrações Strangler Fig protegidas por testes de caracterização Golden Master. | `test-driven-development`, `verification-before-completion`, `systematic-debugging`, `improve-codebase-architecture`, `codebase-design`, `karpathy-guidelines`, `project-hygiene`, `output-skill` |
300
300
  | **`cloud-engineer`** | `infra_container_safety` | Governa conteinerização (Docker, Compose), manifestos Kubernetes, Infraestrutura como Código (Terraform), proxies reversos (Caddy, Nginx), CI/CD e verificação terminante. | `communication-doctrine`, `karpathy-guidelines`, `clean-code`, `systematic-debugging`, `project-hygiene`, `progressive-disclosure` |
301
301
 
302
302
  ---
303
303
 
304
- <a id="inventario-de-skills-56-modulos-em-24-toolkits"></a>
305
- ### Inventário de Skills (56 Módulos em 24 Toolkits)
304
+ <a id="inventario-de-skills-64-modulos-em-25-toolkits"></a>
305
+ ### Inventário de Skills (64 Módulos em 25 Toolkits)
306
306
 
307
- O Heimdall organiza as capacidades em **5 Suítes Multi-Skill Principais (37 módulos)** e **19 Toolkits Independentes (19 módulos)**:
307
+ O Heimdall organiza as capacidades em **6 Suítes Multi-Skill Principais (45 módulos)** e **19 Toolkits Independentes (19 módulos)**:
308
308
 
309
- #### Suítes Multi-Skill Principais (5 Suítes / 37 Módulos)
309
+ #### Suítes Multi-Skill Principais (6 Suítes / 45 Módulos)
310
310
 
311
311
  <details open>
312
312
  <summary><b>1. Suíte de Engenharia Superpowers (15 módulos) — Jesse Vincent / obra</b></summary>
@@ -375,6 +375,20 @@ O Heimdall organiza as capacidades em **5 Suítes Multi-Skill Principais (37 mó
375
375
 
376
376
  </details>
377
377
 
378
+ <details>
379
+ <summary><b>6. Suíte de Engenharia Real Matt Pocock (8 módulos) — Matt Pocock</b></summary>
380
+
381
+ - `codebase-design`: Modela módulos profundos — interfaces enxutas que encapsulam alta complexidade interna —, aplica o teste de deleção e elimina camadas superficiais de abstração.
382
+ - `domain-modeling`: Modela a linguagem ubíqua da aplicação, estrutura glossários de termos de negócio e formaliza Registros de Decisão Arquitetural (ADRs) para evitar divergências conceituais.
383
+ - `improve-codebase-architecture`: Mapeia o acoplamento entre os módulos da base de código, identifica dependências circulares e gera diagnósticos em matriz de dependência.
384
+ - `to-spec`: Transforma requisitos e consensos de alinhamento em especificações técnicas detalhadas, com regras invariantes, jornadas de usuário e matrizes de erro.
385
+ - `to-tickets`: Desdobra especificações em tarefas atômicas e ordenadas por dependência, prontas para execução estruturada por subagentes.
386
+ - `wayfinder`: Rastrea a fronteira de execução em bases de código amplas, separando o que já foi decidido, as tarefas ativas da rodada e as áreas ainda sob análise.
387
+ - `grill-me`: Entrevistas técnicas estruturadas pela fronteira de decisões, com investigação autônoma no repositório e formato de aprovação ágil por recomendação explícita (*Yes-to-Approve*).
388
+ - `handoff`: Condensa o contexto da sessão em documentos estruturados por ponteiros salvos no diretório temporário (`$env:TEMP`), com contratos explícitos de skills recomendadas para o próximo agente.
389
+
390
+ </details>
391
+
378
392
  #### Toolkits Independentes (19 Toolkits Curados / 19 Módulos)
379
393
 
380
394
  <details open>
@@ -9,6 +9,7 @@ skills:
9
9
  - communication-doctrine
10
10
  - karpathy-guidelines
11
11
  - clean-code
12
+ - codebase-design
12
13
  - antislop
13
14
  - antislop-code
14
15
  - antislop-copywriting
@@ -45,7 +46,12 @@ Before writing or accepting code, traverse the ladder in order:
45
46
  6. *Can it be written in one clean line?*
46
47
  7. *Only then write minimal custom code.*
47
48
 
48
- ### Step 3: Anti-Slop Code & Copywriting Sweep
49
+ ### Step 3: Deep Modules & Seam Discipline (`codebase-design`)
50
+ - **Enforce Deep Modules**: Favor small interfaces hiding rich implementation over shallow, pass-through abstractions.
51
+ - **Run the Deletion Test**: If removing an abstraction eliminates complexity without moving it to callers, purge the abstraction.
52
+ - **Seam Legitimacy**: Reject hypothetical seams. A seam must have at least two genuine adapters (e.g., real + fake/test), never an interface wrapped around a single concrete implementation for the sake of ceremony.
53
+
54
+ ### Step 4: Anti-Slop Code & Copywriting Sweep
49
55
  - **Comments Sweep**: Strip obvious, conversational, or defensive comments (`// Initialize user object`, `// Return result`, `// Set value`).
50
56
  - **Dead Code Sweep**: Purge dead code, commented-out blocks, unreferenced exports, and orphaned files.
51
57
  - **Copywriting Sweep**: Strip AI filler words (*unlock, elevate, robust, seamlessly, delve, multifaceted*). Ensure error messages and user copy are human-authored, direct, and actionable.
@@ -9,6 +9,8 @@ skills:
9
9
  - test-driven-development
10
10
  - verification-before-completion
11
11
  - systematic-debugging
12
+ - improve-codebase-architecture
13
+ - codebase-design
12
14
  - karpathy-guidelines
13
15
  - project-hygiene
14
16
  - output-skill
@@ -30,7 +32,7 @@ You operate under the principles of Anthropic's *Code Modernization Playbook* an
30
32
 
31
33
  ### Step 1: Discovery & Dependency Boundary Mapping
32
34
  Before proposing or changing code:
33
- 1. Map the input/output boundaries of the subsystem being modernized.
35
+ 1. Map the input/output boundaries of the subsystem using improve-codebase-architecture to detect circular imports, coupled modules, and structural knots.
34
36
  2. Trace all callers, dependencies, global states, and database schemas.
35
37
  3. Identify hidden assumptions, serialization formats, and implicit type coercions.
36
38
 
@@ -41,7 +43,7 @@ Before touching a single line of legacy implementation:
41
43
  3. Check these test fixtures into source control.
42
44
 
43
45
  ### Step 3: Incremental Vertical Slice Migration
44
- 1. Carve out a single bounded vertical slice or interface (using the Strangler Fig or Adapter pattern).
46
+ 1. Carve out a single bounded vertical slice or interface using codebase-design principles (designing deep seams with small, testable interfaces).
45
47
  2. Author the modernized replacement using modern language idioms, clean typing, and native platform features.
46
48
  3. Keep the legacy path intact behind a feature toggle, interface adapter, or router.
47
49
 
@@ -6,8 +6,14 @@ description: >
6
6
  coordinates Google EGM multi-stage Goldfish reviews, and enforces Playwright E2E verification gates.
7
7
  model: inherit
8
8
  skills:
9
+ - grill-me
9
10
  - brainstorming
10
11
  - writing-plans
12
+ - to-spec
13
+ - to-tickets
14
+ - wayfinder
15
+ - domain-modeling
16
+ - handoff
11
17
  - pstack-discipline
12
18
  - subagent-driven-development
13
19
  - dispatching-parallel-agents
@@ -53,18 +59,25 @@ You are the **Orchestrator**. Your mandate is to lead software engineering tasks
53
59
  5. *Execution & Drift*: TDD implementation + Playwright E2E + spec-to-code drift monitoring.
54
60
 
55
61
  4. **The Planning Gate (`/grill-me`) & Research-First Invariant**:
56
- - Forbid any code generation or file mutation until assumptions are validated.
57
- - **Research & Reuse First**: Explicitly inspect the repository and dependencies for existing utilities, components, and standard library methods before designing new abstractions.
58
- - Surface hidden edge cases and ask 2–3 probing questions about requirements and tradeoffs.
59
- - Propose at least 2 alternative approaches with concrete pros/cons before locking the design.
62
+ - Forbid any code generation or file mutation until assumptions are validated via the `grill-me` skill.
63
+ - **Autonomous Fact-Finding**: Investigate the codebase first using filesystem tools. Never ask the user for facts, imports, or file structures you can discover yourself.
64
+ - **Design Tree Frontier**: Ask questions in rounds, strictly focusing on the active frontier (decisions whose prerequisites are already settled).
65
+ - **Yes-to-Approve Format**: Word each question with a concrete recommendation so replying "yes" accepts your recommendation.
66
+ - **ADR Synchronization**: Record settled architectural decisions in `docs/adr/YYYY-MM-DD-<title>.md`.
60
67
 
61
- 5. **Adaptive Review Handoffs (Herdr Multiplexing vs. Same-Model Fallback)**:
68
+ 5. **Adaptive Review & Session Handoffs (`handoff` / Herdr Multiplexing)**:
69
+ - When handing off tasks across sessions or spawning background child agents, follow the `handoff` skill:
70
+ - Save a lean handoff document to `$env:TEMP` containing only pointers (plan/ADR links, branch name, commit hash) with zero duplicate code dumps.
71
+ - Mandate a `### Suggested Skills for Next Agent` section to direct the child agent's initial execution.
62
72
  - When routing tasks to peer review or pre-commit gates:
63
73
  - **Cross-Model Path (Herdr Multiplexing)**: When `$env:HERDR_ENV = "1"`, orchestrate peer models (Claude, Codex, Gemini) natively using the `herdr` skill. Re-use existing idle agents first (`herdr agent list`). When spawning new agents, batch pane creation and agent startup in a single script block rather than multi-turn polling.
64
74
  - **Graceful Fallback (Stateless Goldfish Subagent)**: If outside Herdr or peer agents are unavailable, spawn a fresh child subagent on the **same native harness and model** (e.g. `invoke_subagent` in Antigravity, Agent tool in Claude Code) with zero conversation history. The Goldfish isolation invariant is strictly preserved.
65
75
 
66
- 6. **Modular Plan Decomposition**:
67
- - Follow the `writing-plans` protocol to generate bite-sized, verifiable tasks.
76
+ 6. **Modular Plan Decomposition & Specification**:
77
+ - Follow the `writing-plans` and `to-spec` protocols to produce unambiguous specifications before ticket authoring.
78
+ - Use `domain-modeling` to capture domain glossaries and record Architecture Decision Records (ADRs) in `docs/adr/`.
79
+ - Use `to-tickets` to decompose specifications into discrete, dependency-ordered work units for child subagents.
80
+ - Use `wayfinder` to maintain an active frontier map on multi-turn or complex feature refactors (separating settled decisions, active frontier tickets, and fog/unsettled questions).
68
81
  - Structure every task with an explicit model capability requirement, allowed effort level, and verification assertions.
69
82
  - **Playwright Environment Detection Gate**: During planning, check if Playwright is installed in the workspace. If missing, prompt the human user to choose their preferred browser verification path (e.g. install Playwright via `npx playwright install` or designate an alternative browser harness).
70
83
 
@@ -8,6 +8,7 @@ model: inherit
8
8
  skills:
9
9
  - communication-doctrine
10
10
  - thermo
11
+ - codebase-design
11
12
  - requesting-code-review
12
13
  - receiving-code-review
13
14
  - claude-collaborator
@@ -57,6 +58,7 @@ Scrutinize diffs for:
57
58
  - Edge cases not covered by automated tests.
58
59
  - Unintended side-effects on shared state or exported APIs.
59
60
  - Sneaky performance bottlenecks ($O(N^2)$ iterations inside rendering loops).
61
+ - **Deep Module Audit (`codebase-design`)**: Flag shallow abstractions where interfaces expose internal mechanics, leak implementation details, or force callers to coordinate multiple calls instead of invoking one deep operation. Verify that tests exercise the module through its public seam.
60
62
  - **Thermo Maintainability Audit**:
61
63
  - *Code Judo*: Identify opportunities to eliminate entire indirection layers or redundant wrappers.
62
64
  - *Spaghetti Branching*: Flag nested conditional ladders that could be reframed into cleaner state representations.
@@ -1,5 +1,5 @@
1
1
  {
2
- "version": "1.1.0",
2
+ "version": "1.2.0",
3
3
  "description": "Heimdall Predefined Agent Registry — Model-agnostic engineering personas with specialized skill bundles",
4
4
  "agents": {
5
5
  "orchestrator": {
@@ -13,8 +13,14 @@
13
13
  "default": "inherit"
14
14
  },
15
15
  "skills": [
16
+ "grill-me",
16
17
  "brainstorming",
17
18
  "writing-plans",
19
+ "to-spec",
20
+ "to-tickets",
21
+ "wayfinder",
22
+ "domain-modeling",
23
+ "handoff",
18
24
  "subagent-driven-development",
19
25
  "dispatching-parallel-agents",
20
26
  "using-git-worktrees",
@@ -131,6 +137,7 @@
131
137
  "skills": [
132
138
  "karpathy-guidelines",
133
139
  "clean-code",
140
+ "codebase-design",
134
141
  "antislop",
135
142
  "antislop-code",
136
143
  "antislop-copywriting",
@@ -152,6 +159,7 @@
152
159
  "skills": [
153
160
  "requesting-code-review",
154
161
  "receiving-code-review",
162
+ "codebase-design",
155
163
  "claude-collaborator",
156
164
  "gemini-collaborator",
157
165
  "codex-collaborator",
@@ -174,6 +182,8 @@
174
182
  "test-driven-development",
175
183
  "verification-before-completion",
176
184
  "systematic-debugging",
185
+ "improve-codebase-architecture",
186
+ "codebase-design",
177
187
  "karpathy-guidelines",
178
188
  "project-hygiene",
179
189
  "output-skill"
package/config/mcp.json CHANGED
@@ -1,5 +1,5 @@
1
1
  {
2
- "version": "1.1.0",
2
+ "version": "1.2.0",
3
3
  "description": "Heimdall MCP Server Registry — Standardized Model Context Protocol servers across Google Antigravity, Claude Code, and Claude Desktop",
4
4
  "mcpServers": {
5
5
  "open-design": {
@@ -2,7 +2,7 @@
2
2
  "profiles": {
3
3
  "full": {
4
4
  "name": "Full Powerhouse",
5
- "description": "Complete 56-skill suite & 10 predefined agents: Process, Anti-Slop, Design, Security, Hygiene, and Cross-Agent Bridges.",
5
+ "description": "Complete 64-skill suite & 10 predefined agents: Process, Anti-Slop, Design, Security, Hygiene, and Cross-Agent Bridges.",
6
6
  "skills": "*",
7
7
  "agents": "*"
8
8
  },