DevOps & Infra

serialexperimentslainnnn Skill

Secrets lifecycle for already-generated credentials. Use when working with HashiCorp Vault or OpenBao (policies, dynamic secrets, seal/unseal, vault or bao CLI), Infisical, Bitwarden Secrets Manager, consuming AWS Secrets Manager, Azure Key Vault or GCP Secret Manager, External Secrets Operator (ExternalSecret, ClusterSecretStore), Secrets Store CSI Driver, sealed-secrets, SOPS with age, systemd LoadCredential= and systemd-creds, gitleaks, betterleaks or trufflehog scans, .gitleaks.toml, responding to a leaked credential, rotation runbooks, or replacing a static credential with short-lived federated identity.

serialexperimentslainnnn Skill

Classical (non-deep) machine learning on tabular data as an engineering discipline. Use when deciding whether a model is needed at all instead of a SQL query, a business rule or a heuristic, splitting data with train_test_split, StratifiedKFold, GroupKFold, TimeSeriesSplit or nested cross-validation, hunting data leakage from a scaler fit on the full dataset, a target-encoded column, an ID column or a future timestamp, building a scikit-learn Pipeline and ColumnTransformer so preprocessing is fit inside the fold, training gradient boosting with xgboost, lightgbm, catboost, HistGradientBoostingClassifier or a linear/logistic baseline with statsmodels, choosing metrics with accuracy_score, roc_auc_score, average_precision_score, precision_recall_curve, f1_score, confusion_matrix, calibrating probabilities with CalibratedClassifierCV, brier_score_loss or a reliability diagram, picking a decision threshold as a product decision, resampling with imbalanced-learn SMOTE and its calibration cost, interpreting with fe

serialexperimentslainnnn Skill

Training your own deep neural network as an engineering decision, not a default. Use when justifying a custom network against a classical model, a frozen pretrained backbone or a third-party API, writing PyTorch training code with nn.Module, DataLoader, torch.compile, torch.amp autocast and GradScaler, gradient accumulation and gradient clipping, checkpointing and resuming with torch.save/load_state_dict, seeding with torch.manual_seed and torch.use_deterministic_algorithms, scaling with DistributedDataParallel, FSDP, torchrun or accelerate versus model/tensor/pipeline parallelism, diagnosing a loss that will not go down or a train/val curve gap, dataloader bottlenecks and storage formats (webdataset, Parquet, tfrecord, memory-mapped tensors), label quality and annotation error, holding out an untouched test set, or shrinking a model for deployment with quantization, pruning, distillation, ONNX or TorchScript export. Also covers PyTorch versus JAX versus Keras/TensorFlow selection.

serialexperimentslainnnn Skill

Use when a GPU must be provisioned, shared, monitored or paid for — pinning the NVIDIA driver and CUDA toolkit to a compatibility matrix, nvidia-open versus proprietary kernel modules, DKMS rebuilds after a kernel update, Secure Boot module signing with MOK, blacklisting nouveau, nvidia-container-toolkit and nvidia-ctk with CDI, the Kubernetes device plugin, GPU Operator, DRA driver and nvidia.com/gpu requests, MIG profiles, CUDA MPS and time-slicing, nvidia-smi, nvidia-persistenced, DCGM and dcgm-exporter metrics, XID errors, ECC and thermal or power throttling, GPU TDP, rack density and liquid cooling, Slurm gres scheduling, AMD ROCm and HIP, or GPU utilization as a FinOps metric.

serialexperimentslainnnn Skill

Planning and executing the migration to post-quantum cryptography — the transition, not the PKI. Use when scoping harvest-now-decrypt-later exposure by data lifetime, building a cryptographic inventory or a CBOM (CycloneDX crypto assets), or working with FIPS 203 ML-KEM, FIPS 204 ML-DSA, FIPS 205 SLH-DSA, the pending FIPS 206 FN-DSA and HQC, legacy Kyber/Dilithium/Falcon/SPHINCS+ names, hybrid TLS 1.3 key exchange with X25519MLKEM768 (IANA group 4588 / 0x11EC) or the retired X25519Kyber768Draft00 (0x6399), OpenSSH KexAlgorithms mlkem768x25519-sha256 and sntrup761x25519-sha512@openssh.com, IKEv2 additional key exchanges (RFC 9370, ke1_mlkem768) and RFC 8784 preshared keys, liboqs / oqsprovider / Open Quantum Safe, PQC support in AWS KMS, Google Cloud KMS or Azure Managed HSM, signature and key sizes blowing up a handshake, a firmware image or a certificate chain, crypto-agility as a design requirement, or the migration deadlines in NIST IR 8547, SP 800-131A and NSA CNSA 2.0.

boxlogodev Bundle

This skill handles SAP BTP integration platform tasks across SAP Integration Suite (CPI - Cloud Platform Integration), SAP Datasphere (formerly DWC), Cloud Connector, OData services, API Management, Event Mesh, Open Connectors, iFlow design (REST, SOAP, IDoc, SuccessFactors, S/4 OData), error handling, certificate management, monitoring, message reprocessing, Datasphere Spaces, views, federation, replication, S/4 ABAP CDS exposure, BTP destinations, and Pre-packaged Integration Content. Use whenever the user mentions CPI, Integration Suite, iFlow, Datasphere, DWC, Cloud Connector, API Management, Event Mesh, OData, IDoc cloud, ABAP CDS exposure, or any cloud integration.

serialexperimentslainnnn Skill

Salesforce and configurable business SaaS - governor limits as an architectural constraint, clicks versus code, and licence cost as a design input. Use when working with Apex classes and triggers (.cls, .trigger), Lightning Web Components (.js-meta.xml, lwc/ directories), Visualforce, Flow Builder and .flow-meta.xml, Process Builder or Workflow Rule end of support and Migrate to Flow, SOQL and SOSL, Apex governor limits (100 SOQL, 150 DML, 10000 ms CPU, 6 MB heap) and bulkification, custom objects and __c / __r fields, record types, validation rules, profiles, permission sets and permission set groups, organization-wide defaults, role hierarchy, sharing rules and Apex managed sharing, WITH USER_MODE, WITH SECURITY_ENFORCED, Security.stripInaccessible and without sharing classes, sfdx-project.json, package.xml, Metadata API, Salesforce CLI (sf project deploy), scratch orgs, unlocked and managed packages, sandbox types and refresh intervals, change sets, AppExchange package due diligence and the security review

serialexperimentslainnnn Skill

Computing on nodes you cannot walk up to — fleet operations for edge sites and devices under an intermittent link. Use when deciding whether a workload actually belongs at the edge (latency budget, upstream bandwidth cost, data residency, offline survival) or is just a distributed monolith, designing an A/B dual-partition image update with automatic rollback and a health check gate (rpm-ostree, bootc, greenboot, balenaOS), rolling an update across thousands of nodes in waves with a kill switch, running a lightweight Kubernetes at the edge (k3s, MicroShift, KubeEdge, Akri) or deciding that systemd plus Podman Quadlet units are enough, store-and-forward telemetry, metric downsampling and egress cost per node, eventual reconciliation and conflict resolution after a reconnect, giving each node its own identity instead of one shared fleet credential, UEFI Secure Boot and measured boot on an unattended node, LUKS full-disk encryption where the attacker physically holds the device, zero-touch onboarding and remote a

serialexperimentslainnnn Skill

Geospatial data as an engineering discipline — coordinate reference systems, formats and spatial SQL. Use when working with EPSG codes (EPSG:4326, EPSG:3857, EPSG:25830, EPSG:4258), WGS84 versus ETRS89 versus ITRF datums, map projection choice and distortion, .shp/.shx/.dbf/.prj shapefiles and their 2 GB and 10-character limits, GeoPackage .gpkg, GeoJSON and RFC 7946 CRS84, FlatGeobuf .fgb, GeoTIFF and Cloud Optimized GeoTIFF (OGC 21-026), GeoParquet, Zarr, PMTiles, COPC, STAC catalogs, PostGIS (ST_Intersects, ST_DWithin, ST_IsValid, ST_MakeValid, ST_Transform, ST_SetSRID, ST_Simplify, ST_Area, ST_Buffer, geography versus geometry, GiST and SP-GiST indexes, spatial_ref_sys), SpatiaLite, DuckDB spatial, GDAL/OGR with ogr2ogr, gdalwarp, gdal_translate and gdalinfo, PROJ and proj.db grid shifts, OGC WMS/WFS/WMTS/WCS and the OGC API - Features/Tiles/Maps family, vector and raster tiles, GeoServer, MapServer, pg_tileserv, TiTiler, Leaflet, OpenLayers, MapLibre GL JS versus the Mapbox GL JS licence change, OpenStre

iambedant Skill

Use this skill when the user wants to learn a concept, technology, or system interactively in a Socratic, teacher-student style. Trigger phrases: "teach me about kubernetes", "I want to learn oauth", "explain tcp/ip to me", "learn [topic]", "/learn [topic]". Different from /teach (which is for codebases) — this skill builds the curriculum from the concept's own structure, starting with the problem the technology was invented to solve. One concept at a time, gated by check-in questions, with explicit detour handling and a personal learning-notes document at the end. Works for any technical concept.

serialexperimentslainnnn Skill

Use when provisioning, hardening or rebuilding the machine a developer works on — versioned dotfiles and a bootstrap script, chezmoi/yadm/GNU stow, Homebrew Brewfile and brew bundle, winget import/export and winget configuration, Nix flake.nix, home-manager, devenv.nix and direnv .envrc, mise with mise.toml and .tool-versions, asdf, nvm/pyenv/rbenv/rustup/uv, .devcontainer/devcontainer.json and the devcontainer CLI, GitHub Codespaces or a cloud dev environment, .editorconfig and formatter/linter config committed to the repo instead of the machine, .vscode/settings.json and .vscode/tasks.json from an untrusted repo, VS Code or Open VSX extension supply chain and malicious extension incidents, curl | sh installers, full-disk encryption with FileVault/BitLocker/LUKS, screen lock and MFA on the workstation, SSH keys in hardware with ed25519-sk or ecdsa-sk and resident or verify-required options, gpg.format ssh and a signing key held on a YubiKey, credential.helper store writing plaintext to ~/.git-credentials, to

serialexperimentslainnnn Skill

Multiplayer game hosting infrastructure. Use when orchestrating dedicated game servers with Agones (GameServer, Fleet, FleetAutoscaler CRDs, agones-sdk), running session-based servers on Kubernetes, matchmaking as a service (Open Match / open-match2, matchmaker tickets and backfill), managed backends (Amazon GameLift Servers, Azure PlayFab Multiplayer Servers, Unity Multiplay, Epic Online Services, Nakama/Heroic Labs, Edgegap), fleet scaling and cost per CCU or per session, UDP DDoS protection for game traffic, server browser and session allocation, or in-game voice/chat services and their compliance.

serialexperimentslainnnn Skill

Cloud security posture as a transversal discipline across AWS, Azure and Google Cloud at once — what the per-provider skills cannot answer. Use when deciding what CSPM, CWPP, CIEM and CNAPP actually mean and which problem each one solves, choosing between open tooling (Prowler, ScoutSuite, CloudSploit, Steampipe and Powerpipe mods, Cartography, Cloud Custodian, Checkov, Conftest) and a commercial posture suite, running a CIS Foundations Benchmark assessment across several accounts, subscriptions or projects at once, building a multi-account or multi-tenant baseline and landing-zone guardrails, preferring preventive policy-as-code in the pipeline and in admission over after-the-fact findings, measuring effective permissions versus granted permissions and hunting wildcard or unused entitlements that no vulnerability scanner will ever report, replacing a flat list of findings with attack-path or toxic-combination analysis, inventorying internet-exposed resources, reconciling declared infrastructure against what

serialexperimentslainnnn Skill

Use when data must be trustworthy and owned — naming a data owner and data steward per dataset versus the platform team, federated ownership and data mesh honesty, choosing or operating a data catalog (DataHub, OpenMetadata, Amundsen, Apache Atlas, Unity Catalog OSS, Collibra, Alation, Atlan), technical versus business metadata, column-level lineage and impact analysis, a business glossary where two teams define "active customer" differently, data contracts as schema plus semantics plus SLA plus owner (Open Data Contract Standard, Bitol ODCS/ODPS, datacontract.yaml) and what happens when one breaks, the quality dimensions (completeness, uniqueness, validity, consistency, timeliness, accuracy) turned into executable assertions, where to check them (source, pipeline, consumption), quality tooling (Great Expectations/GX Core, Soda, Elementary, Evidently), severity of a data incident and notifying the consumers who already decided with bad numbers, data classification tiers (public/internal/confidential/restricte

serialexperimentslainnnn Skill

Containers as systemd services on a single host with Podman and Quadlet, without an orchestrator. Use when writing or debugging Quadlet unit files (.container, .pod, .volume, .network, .kube, .build, .image, .artifact) under /etc/containers/systemd or ~/.config/containers/systemd, running quadlet -dryrun or podman quadlet list/install/rm, migrating off the deprecated podman generate systemd, podman auto-update with io.containers.autoupdate=registry and podman-auto-update.timer, netavark and aardvark-dns or pasta rootless networking and default_rootless_network_cmd, containers.conf, storage.conf, registries.conf unqualified-search-registries and policy.json image trust, podman secret versus systemd LoadCredential=, --userns=keep-id, named volumes versus bind mounts, podman.socket with DOCKER_HOST for docker compose, podman-compose, podman system prune housekeeping, or deciding between a single-host Quadlet stack and a real orchestrator.

serialexperimentslainnnn Skill

Container runtime security and container escape defense. Use when writing or debugging seccomp profiles (RuntimeDefault, seccomp.json, --security-opt seccomp), choosing or pinning a container runtime (runc, crun, gVisor/runsc, Kata Containers, RuntimeClass), rootless Podman or user namespaces (hostUsers, /etc/subuid), --privileged, capability drops (CAP_SYS_ADMIN, CAP_SYS_MODULE, CAP_BPF), a mounted docker.sock or containerd.sock, hostPath/hostPID/hostNetwork/hostIPC exposure, runtime detection with Falco rules, Tetragon TracingPolicy, Tracee or KubeArmor, eBPF agent privileges, container drift and read-only rootfs, or forensic container checkpointing with CRIU/checkpointctl and node/runtime log capture.

serialexperimentslainnnn Skill

Use for the technical response to a security compromise and its investigation — NIST SP 800-61r3 and SANS PICERL phases, containment that preserves evidence, RFC 3227 order of volatility, memory and disk imaging, hashing and chain of custody, cloud snapshot and control-plane log acquisition, container and ephemeral artifacts, super-timeline with plaso/log2timeline and Timesketch, Velociraptor, GRR, KAPE, Volatility 3, Autopsy/Sleuth Kit, YARA-X, CyberChef, IOC and ATT&CK mapping, eradication and rebuild from trusted source, mass credential and token rotation, ransomware, identity-compromise, supply-chain and insider playbooks, breach notification timelines under GDPR, NIS2 and DORA.

ontology-of-everything Bundle

Runs concept-spec guardrails in audit, concept, drift, pipeline, sync, or map mode only when the user explicitly invokes $concept-guardrails. Use to create, backfill, compare, or map colocated concept specifications.

ontology-of-everything Bundle

Run concept-spec guardrails (audit, concept, drift, pipeline, sync, map) only when the user explicitly invokes $concept-guardrails-cn. / 仅在用户显式调用 $concept-guardrails-cn 时运行概念规格护栏。

ono-sendai-labs Bundle

Review an implementation as a whole — either a single completed step or a finished plan. Looks across every commit in scope, the current state of the codebase, the design, and the task-level review residue to surface architectural drift, duplication, undesirable dependencies, doc/code divergence, and cross-cutting issues that no single task review could catch. Produces a structured YAML report and a set of follow-on remediation `.code-task.md` files that re-enter the existing pipeline. Designed to run with clean context after the last task in scope is committed.

ono-sendai-labs Skill

Orchestrate the structured spec-to-code workflow — a pipeline that takes an idea from rough concept through design, planning, task generation, and implementation. Use when the user mentions "spec to code", wants to start or continue a structured development workflow, or asks where they are in the process. Also use when the user has an idea or project and you need to determine which stage of the pipeline to enter.

atomachinskiy Bundle

Яндекс.Трекер — задачи, очереди, поиск тикетов через REST API. Работает с общим yandex-auth токеном плюс X-Org-Id (для Yandex 360) или X-Cloud-Org-Id (для Cloud). Triggers: yandex tracker, трекер, задачи, очереди, тикеты, issues.

atomachinskiy Bundle

Анализ поискового спроса через Yandex Wordstat. Топ запросов, динамика, региональная статистика. Поддержка двух backend'ов: legacy OAuth (для существующих пользователей с одобренной заявкой) и cloud IAM (для новых, через Yandex Cloud Search API). Auto-dispatcher выбирает по конфигу. Triggers: wordstat, вордстат, поисковый спрос, частотность запросов, семантическое ядро, сезонность.

ugaritco Bundle

Deploys and manages Ugarit applications on Ugarit Cloud using the `cloud` CLI. Use when the user wants to deploy an app, ship to Cloud, create or manage applications, environments, databases, caches, object storage, queues, domains, instances, background processes, secrets, compute, scheduled tasks, or Ugarit Cloud infrastructure; or check billing, usage, or spend. Triggers on deployment, shipping, Cloud management, environment setup, database provisioning, billing or usage queries, the `cloud` CLI, and Ugarit Cloud deployment troubleshooting.

kiungsong Bundle

Use when an agent needs structured market, macro, portfolio, calendar, fundamentals, valuation (DCF / reverse DCF / S&P 500 DCF, including turnaround mode), SEC filings, sentiment/breadth, or watchlist research through TerraFin's optimized processing pipeline, including progressive history metadata and optional chart opening.

rmazrim Skill

Validate and lint Kubernetes manifest files (YAML/JSON) against required fields, resource constraints, and best practices, with an optional kubectl dry-run schema check.

rmazrim Skill

Flexible orchestration engine that executes pipeline workflows in Full Mode (all 7 steps sequentially) or Selective Mode (user-specified step subset). Performs data contracting and validation between active steps, skips inactive steps gracefully without breaking the chain, and halts execution with exact error logging when an active step fails.

kelinchen Bundle

自动化 siRNA 药物设计流水线（16 步）——siDirect 设计、热力学/可及性、miRNA seed、脱靶 BLAST、基因组坐标、gnomAD 变异、家族保守性、多物种脱靶、21 参数评分、对标已上市药（Inclisiran/Plozasiran 等）。Use when the user asks to design siRNA, run siRNA design pipeline, analyze siRNA off-targets, knock down a gene with siRNA, compare candidates with approved siRNA drugs, or mentions siDirect / siRNA 设计 / 脱靶 / Inclisiran. Works standalone via CLI on Windows/macOS/Linux.

rmazrim Skill

Design automated CI/CD workflows for testing, building, and deployment.

rmazrim Skill

Draft modular, reusable Infrastructure-as-Code Terraform modules for cloud infrastructure.

tiennguyen1203 Skill

Scaffold a new RegASK backend workload end-to-end — source repo cloned from the fleet's newest baseline with the donor's domain stripped, CI/CD wired, and platform-gitops manifests for all four environments. Handles all three app types (service, ai-service, kservice). Use when the user asks to set up / create / scaffold a new service, clone an existing service as a boilerplate, or add a service to platform-gitops. Vietnamese triggers: setup source cho service mới, clone từ thằng X, dựng service mới, thêm service vào gitops.

professor-gpt Bundle

Senior DevOps engineer that designs CI/CD pipelines, container platforms, and infrastructure-as-code with battle-tested deployment strategies and DORA-metric discipline.

rmazrim Skill

Orchestrates the 5-stage infrastructure and QA chain — ORM schemas plus zero-downtime migrations (skills/drizzle-prisma-orm-architect), Redis caching and pub/sub event bus (skills/redis-pubsub-cache-manager), multi-stage Docker images with Compose stack (skills/docker-multi-stage-stack-builder), Playwright E2E and security flows (skills/playwright-e2e-security-flow-tester), and Lighthouse Core Web Vitals audits (skills/lighthouse-web-vitals-optimizer) — all executed against the provisioned stack. Accepts full_run, include_steps, and skip_steps parameters to select the chain subset, applies predecessor gating and fail_fast semantics, captures per-step stdout and reports, and consolidates DB migration status, cache health, docker build/health, E2E pass counts, a Lighthouse scores table, and recommendations into a single INFRA_QA_SUITE_REPORT.md.

rmazrim Skill

Loads Python pyc bytecode via marshal after importlib header validation, pretty-prints raw dis, reconstructs a basic-block control-flow graph annotated with loop heads and try/except regions, renames mangled single-letter variables using usage heuristics, renders a readable structured pseudo-Python outline, flags obfuscation signatures such as missing strings, oversized constants and dynamic eval execution, and documents the uncompyle6 byte-exact pipeline when it is installable.

rmazrim Skill

Scaffold lightweight serverless handlers for Cloudflare Workers or AWS Lambda with standard CORS and error handling.

rmazrim Skill

Offline-first subdomain enumeration and service fingerprinting using only the Python standard library. Brute-force hostname resolution via socket for live targets with clean text and JSON reports; offline mode replays a resolv-host map so the full pipeline runs in an air-gapped sandbox. No external DNS tools, no MCP.