DevOps & Infra

davidcastagnetoa Bundle

GitOps — el estado del clúster Kubernetes refleja siempre el estado del repositorio Git

davidcastagnetoa Bundle

Modelo de embeddings faciales alternativo a ArcFace para mayor resiliencia del pipeline

davidcastagnetoa Bundle

Orquestación de contenedores en producción con Helm para gestionar releases versionadas

davidcastagnetoa Bundle

Correlación de logs entre microservicios del pipeline KYC usando trace_id y session_id.

davidcastagnetoa Bundle

Agentes de recolección de logs (Promtail o Vector) para centralizar logs del pipeline KYC en Grafana Loki.

davidcastagnetoa Bundle

Patrones de concurrencia async para ejecutar agentes en paralelo dentro del pipeline KYC

davidcastagnetoa Bundle

Health checks profundos que verifican funcionalidad real de cada componente del pipeline KYC

davidcastagnetoa Bundle

Fallback cloud OCR para casos donde el OCR self-hosted falla

davidcastagnetoa Bundle

Exportar modelos PyTorch del pipeline KYC a formato ONNX para portabilidad y optimización

davidcastagnetoa Bundle

SDK de OpenTelemetry para instrumentación unificada de métricas, trazas y logs del pipeline KYC en Python.

davidcastagnetoa Bundle

Cliente Python de Prometheus para instrumentar el pipeline KYC con metricas custom

davidcastagnetoa Skill

Backup y restore de PostgreSQL con pgBackRest para proteger los datos del pipeline KYC

davidcastagnetoa Bundle

Monitorización de queries lentas y optimización de rendimiento del pipeline KYC

davidcastagnetoa Bundle

Stack de métricas y dashboards para visibilidad del pipeline KYC en tiempo real

davidcastagnetoa Bundle

Pipeline de CI/CD con gates de calidad automatizados en cada PR

davidcastagnetoa Bundle

Supervisor de procesos que reinicia workers muertos, detecta zombies y escala segun carga del pipeline KYC

ascend-ai-coding Bundle

昇腾 NPU 平台 vLLM 大模型推理服务一键部署。触发：用户说'部署 模型名'、'NPU 部署模型'、'vllm serve'。流程：SSH检查 → NPU检查 → 配置发现(必须验证) → 用户确认 → 部署 → cron监控 → 验证。约束：(1) 配置必须从官方文档验证，禁止猜测；(2) 后台启动必须用cron监控，禁止手动轮询。支持 Qwen/Qwen3.5、GLM、DeepSeek、Kimi。

ascend-ai-coding Skill

昇腾NPU CANN Toolkit+Kernels+NNAL安装部署技能。支持从官网下载run包安装和从Docker镜像提取两种方式，覆盖驱动检查、包下载、安装、环境变量配置与验证全流程。当用户需要安装CANN全套组件或指定版本CANN到自定义路径时调用。

ascend-ai-coding Bundle

Verl 分布式训练服务一键拉起与配置。触发场景：(1) 用户要启动 Verl 训练任务或部署 RLHF/DAPO 训练环境 (2) 在 NPU 集群上拉起 Verl 训练容器 (3) 配置 Ray 集群和 SwanLab 监控 (4) 根据 7 位二进制掩码灵活配置加速特性。支持 Qwen3-8B 等 Megatron 模型的 DAPO 训练全流程。

ascend-ai-coding Skill

编译 ATB (Ascend Transformer Boost) 测试框架。当用户需要编译 ATB 测试框架、 运行 CSV 测试、或构建 atb_test_framework 时调用。支持全量编译（含第三方依赖克隆与源替换） 和增量编译两种模式。需在 Docker 容器内配合 CANN 环境执行。

ascend-ai-coding Bundle

MindSpeed-MM skill router and model index for Huawei Ascend NPU. Use when the user is uncertain which MindSpeed-MM skill to use, needs to choose between understanding/generative/omni/audio model categories, or wants an overview of the full training pipeline. Routes to the appropriate leaf skill based on model type.

ascend-ai-coding Bundle

Analyze and debug upstream vLLM test failures on Ascend NPUs. Adapt test cases from `vllm/tests/` for the vllm-ascend plugin, and identify tests that are compatible with the `vllm-ascend` continuous integration (CI) pipeline. This skill should be used to analyze whether upstream vLLM mainline test cases can run successfully on Ascend NPUs, to port upstream tests, to perform compatibility checks, to triage CI tests, and to validate mainline compatibility on the Ascend hardware.

ascend-ai-coding Bundle

MindSpeed-LLM 端到端训练部署流水线，用于华为昇腾 NPU。串联环境搭建、数据预处理、权重转换、分布式训练、推理验证、评估、权重导出的完整流程。覆盖参数一致性校验、阶段间数据流转、模型特定配置（Qwen、LLaMA、DeepSeek）。用户需要在昇腾 NPU 上完成 MindSpeed-LLM 训练部署全流程时使用。

ascend-ai-coding Bundle

Ascend NPU model migration suite. Invoke when user wants to migrate/train models on Ascend NPU, setup environment, or deploy models from open-source repositories.

ascend-ai-coding Bundle

Diffusers Pipeline 推理指南，用于华为昇腾 NPU。覆盖环境预检、通用 Pipeline 推理（图像/视频模型）、内存优化（CPU offload、attention slicing、VAE slicing）、LoRA 加载与融合、多卡推理和按版本检索 Diffusers API。用户一旦提到在昇腾 NPU 上运行 FLUX、SDXL、Wan、CogVideoX 等 Diffusers 推理，或需要排查推理参数、显存优化、LoRA 接入、Pipeline API 路径时，就应使用此 skill。

delorenj Skill

Autonomous audit-to-fix pipeline — find issues, classify, fix, test, commit

ferroxlabs Skill

Entry point for people-operations work at the two-to-twenty-employee stage: collects the jurisdiction, headcount and classification facts that decide which employment rules apply, then routes or works inline — job descriptions, interview loops, onboarding, reviews, 1-on-1s, offer letters, comp bands and performance plans. Use when a people question arrives without a document attached to it. Do NOT use when the task is already named — go straight to hr-handbook, hr-termination-letter, hr-offboard, hr-leave-of-absence, hr-accommodation-request or hr-rif — or when the work is role and hiring design (use slate-role-design). Templates only, never employment-law advice: outputs need review by HR counsel.

ferroxlabs Skill

Develops three named scenarios (optimistic, base, pessimistic) for a decision or situation. Each scenario includes driving assumptions, key events, outcome description, and preparation actions. Produces a complete scenario set, not a description of scenario methodology. Use when the user asks about planning for different outcomes, thinking through best-case and worst-case scenarios, preparing for uncertainty, or building contingency plans. Do NOT use for business strategic scenario planning (use business strategy skills), risk registers for projects (use risk-assessment), or simple pros and cons comparisons (use pro-con-analysis).

ferroxlabs Skill

Strategic thinking frameworks including scenario planning, game theory basics (Nash equilibrium, prisoner's dilemma), long-term thinking, second and third-order effects analysis, strategic frameworks (Blue Ocean Strategy, Wardley Maps), decision journals, pre-mortem analysis, and competitive strategy. Use when the user asks about strategic thinker or needs help with related topics. Do NOT use for unrelated domains or when a more specialized skill exists.

ferroxlabs Skill

CI/CD pipeline design. Pipeline stages, build optimization, test parallelization, deployment strategies (blue-green, canary, rolling), artifact management, environment promotion, pipeline-as-code, security scanning in CI. Use when the user asks about cicd architect, cicd architect best practices, or needs guidance on cicd architect implementation. Do NOT use when the user needs a different specialized skill or is asking about an unrelated technology domain.

ferroxlabs Skill

The user says "the pipeline number is wrong," "the forecast keeps missing," "our CRM is a graveyard," or "I can't trust the data." Load for pipeline audit, contact-data cleanup, stale-deal sweep, or a defensible forecast. Pairs with operating-rhythm — the CRM feeds the weekly tactical's pipeline-cov

ferroxlabs Skill

Advanced monorepo architecture covering Nx, Turborepo, and Lerna tooling, workspace management patterns, build caching strategies, dependency graph optimization, task pipeline design, and migration planning for large-scale codebases. Use when the user asks about monorepo architect, monorepo architect best practices, or needs guidance on monorepo architect implementation. Do NOT use when the user needs a different specialized skill or is asking about an unrelated technology domain.

ferroxlabs Skill

The founder says "I'm stuck on a call," "I keep going back and forth," or "what would you do here?" Load for any binary decision or unresolved tradeoff older than two weeks. If they ask you to decide for them, push back: the frame is yours; the call is theirs.

ferroxlabs Skill

The founder says "my weeks just disappear," "I never have time for the real work," or "I'm always reactive." Load for personal-rhythm questions, weekly-review design, or deep-work installation. If they ask for a company operating rhythm, route to Ops — that's the team layer.

ferroxlabs Skill

A curated collection of essential mental models for better thinking and decision-making, including inversion, second-order thinking, Occam's razor, Hanlon's razor, circle of competence, map vs territory, opportunity cost, margin of safety, via negativa, Lindy effect, and antifragility, with guidance on when to apply each model. Use when the user asks about mental model toolkit or needs help with related topics. Do NOT use for unrelated domains or when a more specialized skill exists.

ferroxlabs Skill

Applies second-order thinking to a decision by mapping direct consequences, then the consequences of those consequences, then third-order effects. Surfaces non-obvious risks and opportunities before the decision is made. Use when the user asks about thinking through consequences, considering ripple effects, understanding downstream impacts, or applying second-order thinking to a decision. Do NOT use for imagining failure scenarios (use premortem-analysis), comparing options with scoring (use weighted-decision-matrix), or business strategic impact analysis (use business strategy skills).