Cloud provider operations: GCP IAM inheritance and impersonation; AWS cost investigation (Cost Explorer + CUR, anomaly detection, Savings Plans vs RIs); and Cloudflare platform — DNS zones (REST), Workers configuration, the cf unified CLI (preview), Zero Trust Access for remote MCP servers, and traffic analytics via the GraphQL Analytics API (per-hostname breakdowns, dataset limits, sampling). Ships stdlib validators.
draw.io diagram authoring: native .drawio file generation, the draw.io MCP servers, and an architecture/HLA method for infrastructure & Kubernetes topology (semantic edge routing, comb-bus/ladder/rail layouts, icon and color conventions). Ships a reusable Python generator.
Install Speedscale end to end, record traffic with proxymock, run and monitor replays, tune the mocks and the tests until a replay is accurate and fully mocked, analyze replay reports, and gate changes with regression, verify-fix, contract, chaos, load and performance tests over recorded traffic
Polylane skills, commands, and MCP servers: investigate production issues with live logs, metrics, traces, deployments, and the cloud infrastructure context graph
cohesivity.ai offers free agent native backend services. Annonymous account (no-signup) to get started through MCP or API. Hosting, postgres, email, storage, containers, LLMs, voice and third-party APIs. Includes free tiers and 5 USD/mo in AI and Search credits. topups through x402.
For whoever runs the gateway: /mtg-gateway-operator:deploy walks through the Swarm, Authentik, Portainer and proxy setup with secrets created as Docker secrets; /mtg-gateway-operator:invite adds a user.
Four Claude Code hooks, Redis inboxes, and a tmux daemon that wakes stopped sessions with pointer prompts while active sessions receive full messages through additionalContext.
development flow for any workbench tree (metarepo or flat): one approval gate over file-based epic tickets and research skills, then epic-run in the epic's integration worktree with one implementer and one reviewer per task and one scope audit per epic, gated review and debugging flows, verification hooks that tie every test run to the committed HEAD, read-only diagnostic subagents, and destructive-command guardrails.
Skills for keeping a running system alive: triage a red pipeline to its real cause, get a broken Kubernetes workload back before diagnosing it, design alerts where every page is actionable, instrument a service so the next incident is answerable, write the runbook a stranger can follow at 3am, rehearse failure on purpose, plan for load before it arrives, and find out why the bill grew.