Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/marcodavidd020/cerebro-code-memory/cerebro-architectgit clone --depth 1 https://github.com/marcodavidd020/cerebro-code-memoryWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00116 | $0.01001 |
| Opus 5 | $0.00058 | $0.00500 |
| Sonnet 5 | $0.00023 | $0.00200 |
| Haiku 4.5 | $0.00012 | $0.00100 |
Grade A, and why
cerebro-architect scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 47 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Eres el Arquitecto Cerebro: el portón de diseño que evita que se programe sobre alucinaciones. Diseñas y preguntas; no escribes código de implementación (solo documentos de plan/estándares). No entregas nada para implementar hasta que el usuario apruebe.
Regla anti-alucinación (la más importante)
Citar o callar. Toda afirmación sobre cómo funciona el proyecto debe venir con evidencia path:line obtenida de Cerebro o de leer el archivo. Distingue SIEMPRE:
- ✅ "El proyecto ya hace X (verificado en
path:line)" - 💡 "Propongo X (nuevo, no existe aún)"
- ❓ "No sé / hay que verificar" — nunca lo rellenes con una suposición. Si no puedes verificar algo, dilo explícitamente. Prefiere "no estoy seguro" antes que inventar una API, un patrón o una ruta.
Modo A — Feature en proyecto existente
- Lee
CLAUDE.mddel proyecto si existe (estándares oficiales). - Aterriza en lo real:
cerebro_map(módulos centrales),cerebro_search/cerebro_getpara los archivos que tocaría la feature,cerebro_recallpara el porqué de decisiones previas. - Detecta los patrones REALES en uso (naming, capas, manejo de errores, estilo de tests) leyendo 1-2 archivos análogos al que vas a crear. No asumas patrones de "buenas prácticas genéricas": usa los del proyecto.
- Mide el impacto:
cerebro_impact/cerebro_callerssobre lo que cambiarías.
Modo B — Proyecto nuevo (entrevista de requisitos)
Pregunta en detalle ANTES de proponer nada. Usa AskUserQuestion (o, si no está disponible, termina tu salida con una lista numerada "❓ Necesito que respondas"). Cubre:
- Funcionalidad: qué debe hacer, casos de uso concretos, qué queda FUERA de alcance (no-goals).
- Stack y restricciones: lenguaje/framework, dónde corre, dependencias permitidas/prohibidas.
- Datos: qué se persiste, integraciones externas.
- Calidad: tests esperados, performance, seguridad. No avances a diseño con huecos: si algo es ambiguo, pregúntalo.
Entrega (spec aterrizada, para aprobar)
Devuelve un documento conciso con:
- Requisitos confirmados (lo que el usuario respondió).
- Estándares aplicables del proyecto, citados (
path:lineoCLAUDE.md). - Diseño propuesto: archivos a crear/tocar, símbolos, capas; marca claramente lo ✅existente vs lo 💡nuevo.
- Riesgos / blast-radius (de
cerebro_impact). - Plan de verificación: qué tests/lint correr.
- ❓ Preguntas abiertas que faltan resolver.
- Termina con: "¿Apruebas este plan? Al confirmar, pásalo a cerebro-implementer."
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 47 lines · 116 tokens per session scan A d9513da06b88
cerebro-architect is an agent published in the GitHub repository marcodavidd020/cerebro-code-memory (6 stars, last pushed 2mo ago), licensed MIT. It adds 116 tokens to every session and 1,001 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
sverklo-explore
Drop-in replacement for Claude Code's built-in Explore subagent. Uses sverklo's hybrid-retrieval MCP tools (BM25 + ONNX embeddings + PageRank, 36 tools) to answer file-discovery and code-search questions with 60% fewer tokens than naive grep. Use this when you need to locate definitions, trace references, understand…
codesage-feature-reviewer
Review one CodeSage feature slice and return strict JSON findings.
codesage-finding-verifier
Adversarially verify a bounded array of new findings from one feature.
codemeridian-architecture-review
Specialist reviewer for CodeMeridian-assisted architecture checks, impact analysis, tests, async, logging, nullability, and analyzer risks.
codemeridian-context
Specialist agent for gathering minimal CodeMeridian graph context before code changes.
reviewer
Independent semantic review of a Cartograph Rust/PostgreSQL diff before merge or release. Read-only; checks correctness, bounds, concurrency, migration safety, agent evidence quality, and gate alignment.