frontier-cockpit

Decision framework for designing production agentic systems on GitHub platform and Azure AI Foundry. Covers model routing tiers (mini to frontier), prompt caching and semantic caching, short and long term memory, context curation and RAG, tools and MCP, identity and guardrails, evaluation, observability (OpenTelemetry GenAI), and cost control. Use when architecting an AI agent or multi-agent system, choosing a model router, deciding a caching or memory strategy, planning context windows, or reviewing an agent design for reliability, security, and cost. Pairs with azure-managed-redis-cache, foundry-agent-blueprint, azure-api-center, apim-ai-gateway, and azure-architecture-diagrams.

Front model and tool backends with Azure API Management as an AI gateway: token rate limiting (token-per-minute and quota), multi-backend load balancing and circuit breaker across model deployments, semantic caching, token metric emission for cost attribution, managed identity to backends, and content safety integration. Use when designing the runtime gateway for an agentic platform, enforcing token budgets, load balancing across model endpoints, adding a semantic cache at the gateway, or governing tool API calls. Routes to the installed azure-aigateway skill for policy detail. Pairs with azure-api-center (governance), azure-managed-redis-cache (cache backend), and agentic-architecture-patterns.

Validate an architecture document (the {app}_Architecture.md produced by the Senior Cloud Architect agent) against a hard Definition of Done: required sections, the five required diagrams, well-formed Mermaid, the seven explanation parts per diagram, and the repository copy conventions. Use when producing, reviewing, or finishing an architecture document, or whenever you need to confirm a Mermaid-based architecture deliverable is complete and will render before presenting it.

Govern APIs and agent tools at enterprise scale with Azure API Center: a central catalog of APIs, versions, definitions (OpenAPI), environments, and deployments, plus registration of agent tools and MCP servers for discovery and reuse. Use when designing API governance for an agentic platform, building a tool or MCP registry for agents, enforcing API standards with the API Center linter, or making backend APIs discoverable to agent builders. Pairs with apim-ai-gateway (runtime gateway), foundry-agent-blueprint (tool consumption), and agentic-architecture-patterns (tools and MCP).

Produce complete, professional architecture diagrams that use the official Azure, Microsoft, and GitHub (Octicons) icon sets, output as editable draw.io (.drawio) files and exported SVG. Covers the official icon catalogs and their terms of use, the draw.io mxGraph file format, layout and connector conventions, and a bundled draw.io MCP server (Python FastMCP) that creates and edits .drawio diagrams programmatically (add nodes with official icons, connect edges, group, and lay out). Use whenever the deliverable includes an architecture diagram, a system context, a component or deployment diagram, or any cloud diagram that must use official vendor icons. Pairs with ai-native-engineer and agentic-architecture-patterns, and complements svg-professional for hand-crafted SVG.

Design and provision Azure Managed Redis as the cache, semantic cache, vector store, session store, and agent memory backend for AI-native systems. Covers SKU selection (Balanced, MemoryOptimized, ComputeOptimized, FlashOptimized), the fact that Redis Enterprise is retired for new creations, vector search and RedisVL for semantic caching and long term memory, session and short term state, tenant isolation, managed identity (Entra) access, private networking, and a Bicep sample. Use when an agent design needs caching, a semantic cache, a vector memory store, or session state on Redis. Pairs with agentic-architecture-patterns and apim-ai-gateway.

Design and provision agents on Azure AI Foundry Agent Service: model catalog and deployments, connections (Azure AI Search, Azure Managed Redis, Bing, storage), threads and runs for short term memory, tools (function, OpenAPI, MCP, code interpreter, file search), evaluation, and tracing. Use when the target runtime is Azure AI Foundry, when choosing models from the Foundry catalog, when wiring agent tools and connections, or when planning Foundry evaluation and observability. Routes to the installed microsoft-foundry, azure-ai, and vscode-microsoft-foundry skills for provisioning detail. Pairs with agentic-architecture-patterns, azure-managed-redis-cache, azure-api-center, and apim-ai-gateway.

Creates professional Markdown documents with YAML frontmatter, versioning, author attribution, table of contents, and consistent formatting. Use for Markdown documents, README, ADR, spec, guide, changelog, runbook, RFC, technical docs, and converting PPTX or PowerPoint files to narrative Markdown reading editions with markitdown. Trigger on 'md file', 'markdown', '.md', 'documento markdown', 'criar um readme', 'escrever documentação', 'write a spec', 'draft a guide', 'convert PPTX to Markdown', 'extract PPTX', or 'PowerPoint to Markdown'. Do not use for creating PPTX presentations, Word documents, diagrams, or PDFs.

Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).

Generates a professional consolidated PDF from multiple Markdown files — playbooks, runbooks, frameworks, or any multi-chapter document. Produces: dark cover page with gradient accent bar and chapter preview blocks, auto-generated TOC with colored dots, per-chapter colored banner headers, continuous page numbers, embedded images, dark-mode code blocks, striped tables, and a 2-column references section on its own final page. Use when the user says: "criar PDF do playbook", "gerar PDF consolidado", "juntar capítulos em PDF", "consolidate chapters into PDF", "make a runbook PDF", "PDF único de vários markdowns", "playbook completo em PDF", "assemble playbook", "build PDF from markdown files", or wants a single polished deliverable from multiple .md files.

Design, improve, and validate prompts, prompt files, custom instructions, custom agents, and skill instructions for GitHub Copilot in VS Code, GitHub Copilot CLI, and GitHub Copilot cloud agent. Use when the user asks to write or improve prompts, create prompt files, design agent instructions, tune context usage, choose task routing, structure tool use, or make GitHub Copilot outputs more reliable.

This skill should be used when the user asks to 'prepare release', 'create PR', 'generate documentation', 'run release gates', or needs guidance on Phase 9 release. Also trigger on 'sdd release', 'blocking gates', 'security scan', 'release gate', or 'export work items'.

Analyze, complement, and validate FRD and NFRD as correct inputs for sdd_init. Use when: creating requirements, FRD, NFRD, analyzing specs, greenfield, brownfield, modernization, migration, API, SaaS, mobile, data platform, internal tool, CLI. Produces FRD (what system must do) and NFRD (quality constraints). DO NOT USE for code, implementation, or CONSTITUTION.md, that is sdd_init's job.

This skill should be used when the user asks to 'research codebase', 'scan tech stack', 'import documents', 'discovery questions', or needs guidance on Phase 1 research. Also trigger on 'sdd research', 'brownfield scan', 'import transcript', or 'check ecosystem'.

Create, audit, repair, and improve GitHub Copilot Agent Skills for VS Code, GitHub Copilot CLI, and GitHub Copilot cloud agent. Use when a user asks to create a skill, generate a SKILL.md, convert a workflow into a reusable skill, audit an existing skill, fix skill loading issues, optimize a skill description, validate frontmatter, or add references/scripts/assets to a skill package.

Create, audit, or consolidate agent skills following the Agent Skills open standard (agentskills.io). Interviews the user relentlessly about intent, scope, and edge cases before drafting. Covers SKILL.md structure, frontmatter, progressive disclosure, description optimization, script bundling, sub-command architecture, setup gates, context systems, and review. Use when the user wants to create a skill, write a skill, build a new skill, make a skill, draft a SKILL.md, or mentions "skill-maker". Also use when asked to review a skill, audit a SKILL.md, check why a skill never triggers, improve an existing skill, or fix a skill. Also use when asked to package expertise, workflows, or domain knowledge into a reusable skill. Also use when asked to consolidate skills, merge skills, combine skills, reduce skill count, or refactor multiple skills into one.

Suggest relevant GitHub Copilot custom agents from github/awesome-copilot for this repository. Use when comparing HQ agents with the Awesome GitHub Copilot catalog, identifying missing agent personas, avoiding duplicate imports, or reviewing whether a local agent should be refreshed from upstream. Produces recommendations only and never installs automatically.

Suggest relevant GitHub Copilot instruction files from github/awesome-copilot for this repository. Use when comparing HQ path-scoped instructions with the Awesome GitHub Copilot catalog, identifying missing coding standards, avoiding duplicate instructions, or planning an adapted instruction import. Produces recommendations only and never installs automatically.

Suggest relevant GitHub Copilot prompt files from github/awesome-copilot for this repository. Use when comparing HQ prompt files with the Awesome GitHub Copilot catalog, identifying missing slash-command workflows, avoiding duplicate prompts, or planning a curated prompt import. Produces recommendations only and never installs automatically.

Suggest relevant GitHub Copilot skills from github/awesome-copilot for this repository. Use when comparing HQ skills with the Awesome GitHub Copilot catalog, finding reusable workflows, avoiding duplicate skills, checking upstream drift, or planning a curated import. Produces recommendations only and never installs automatically.