AI Eng System

Deploy, manage, and troubleshoot Kubernetes workloads. Use for manifest review, Helm chart validation, resource tuning, RBAC, and cluster operations.

Use only when the user explicitly asks for TDD, a failing test, or a regression test, OR when the bug has an obvious cheap local test target. Skip when the test path is unclear, expensive, integration-heavy, or not requested.

Build secure, reliable Slack integrations and bots. Use for webhook design, Block Kit UI, Socket Mode, and API best practices.

Teach the user a topic through a stateful learning workspace — mission-driven lessons, retrieval practice, cited resources, and durable learning records. Use when the user says teach me, /teach, wants interactive lessons, or is learning something over multiple sessions.

Build, run, and secure Docker containers with current best practices. Use for Dockerfile review, multi-stage builds, Compose orchestration, image hardening, and CI/CD integration.

Find failing PR checks, inspect logs or external check links, and apply focused fixes

Operate GitHub repositories, workflows, and PRs efficiently. Use for Actions optimization, PR hygiene, repo maintenance, and team collaboration patterns.

Set up and operate Sentry for error tracking, performance monitoring, and release health. Use for SDK integration, alert tuning, and incident investigation.

Cut AI tells from any writing. Must always apply.

Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when the user wants to stress-test a plan, get grilled on their design, or mentions "grill me".

Sketch types, signatures, and module structure before code, then stay in the loop while implementation fills in. Use for /architect, 'architect this', 'design this', or non-trivial work where jumping to code would lock in the wrong shape.

Relationship-aware retrieval using graph traversal, entity anchors, community expansion, and hybrid vector plus graph search. Use when chunk similarity alone misses paths, entities, or subsystem context.

Adapt existing content into new formats for different channels. Use for "repurpose this", "turn this into", "adapt for", or /repurpose.

Full SEO and performance audit for any website. Use when asked to check SEO, run a lighthouse scan, audit performance, check accessibility, or validate structured data.

Build a local browser/CDP harness to drive and inspect a web, IDE, or Electron UI. Use for UI verification, screenshots, perf profiles, or reproducing UI bugs.

Guide building apps, scripts, CI pipelines, and automations with the Cursor TypeScript SDK (`@cursor/sdk`). Use when integrating `Agent.create`, `Agent.prompt`, `Agent.resume`, streaming, MCP servers, local vs cloud runtime, errors, or porting REST `/v1/agents` calls to the SDK. Prefer this skill over memory—the SDK surface evolves and references here are the source of truth.

Structured document review for structure, clarity, and technical accuracy. Use for "review this doc", feedback, critique, or /doc-review.

Monitor PR checks and fix failures until green. Uses gh pr checks as the source of truth for PR-attached checks.

Design and operate application observability with metrics, logs, traces, and alerts. Use for SLO definition, dashboard design, on-call runbooks, and incident response.

Write, run, and maintain reliable end-to-end tests with Playwright. Use for test generation, CI integration, trace debugging, and cross-browser validation.

This skill should be used when creating extensions for Claude Code or OpenCode, including plugins, commands, agents, skills, and custom tools. Covers both platforms with format specifications, best practices, and the ai-eng-system build system.

Use for "how does X work", code walkthroughs before changing something, and placement / ownership / layering questions ("where should this live", "which package owns this", "is this the right layer"). Explains subsystem architecture, runtime flow, onboarding mental models. Can critique architecture. Use why for motivation.

Use for 'why does X work this way', 'why we picked Y', design rationale, regressions, postmortems, or data-backed thresholds. Discovers available MCPs and queries each evidence category (source control, issue tracker, long-form docs, real-time chat, infrastructure observability, error tracking, product analytics warehouse) in parallel, then returns a cited read on decisions and tradeoffs. Use how for runtime behavior.

Principal-engineer repository audit and prioritized improvement plan. Four phases - discovery, evidence-based audit, strategy, task plan - with file:line citations and severity ratings. Analysis only, no code changes. Use when asked to "audit this repo", "project review", "health check", "improvement plan", or after a model upgrade to re-baseline important projects.

Build, test, and archive iOS/macOS apps with xcodebuild and Xcode CLI tools. Use for CI integration, scheme management, simulator control, and build troubleshooting.

Build a local harness to drive, inspect, and profile an interactive CLI or TUI. Use for CLI UX checks, startup regressions, hangs, or prompt-flow verification.

Transform a vague idea, request, or prompt into a concrete, actionable goal with clear success criteria, scope boundaries, and measurable outcomes. Use when the user has an intention but needs help defining what "done" looks like.

Render a documentation-style Cursor Canvas that organizes architecture notes, API references, walkthroughs, and how-tos into a navigable layout with sections, tables of contents, and cross-references. Use when the user asks for a docs canvas, documentation overview, architecture walkthrough, API reference page, or wants to render structured documentation as an interactive canvas.

Use for "interrogate", "adversarial review", "multi-model review", "challenge this", "stress test this code", "find blind spots", or "tear this apart". Four LLM reviewers challenge changes from independent angles.

Use only when the user explicitly types `/ai-eng/orchestrate <goal>` to decompose a large task, spawn a tree of parallel cloud-agent workers/subplanners/verifiers via the Cursor SDK, and collect structured handoffs; do not invoke autonomously.

Agent evaluation framework. Measure agent performance, identify weaknesses, and track improvement over time. Use when assessing agent quality, comparing approaches, or validating changes.

Manage git worktrees for isolated parallel development. Use when running multiple sessions concurrently or working on parallel features.

Spawn N parallel candidates at the same task, pick a base, graft the strongest parts of the losers into it. Use for /arena, 'arena this', 'throw it in the arena', or when one attempt at a non-trivial artifact would lock in the wrong shape.

Remove AI-generated filler, repetition, inflated language, and redundant comments while preserving meaning.

Daily briefing synthesizing calendar, tasks, and priorities. Use for "morning brief", "daily briefing", "start my day", or /morning-brief.

Multi-source deep research with confidence-rated synthesis. Use for "research deeply", "deep dive on", "comprehensive research", or /research-deep.

Self-continuing loop that fixes a deliberately failing test, reruns it, and exits cleanly when green. Smallest cooking-routines skill — pairs with the cooking Stop hook to keep iterating past natural stops.

Produce a weekly synthesis of authored commits with highlights by bugfix, tech debt, and net-new work

Unified agent application scaffolding across 5 harnesses: Anthropic, Cursor, OpenAI, OpenCode, and Pi. Interactive project generator with live SDK docs, per-harness code patterns, and built-in verification. Use when someone says "create an agent app", "new SDK project", "build with [harness] SDK", "scaffold agent", or wants to verify an existing agent setup.

Design or review CLIs so coding agents can run them reliably: non-interactive flags, fast actionable errors, idempotency, dry-run, layered --help. Use when building a CLI or adding commands.

Deploy applications to Coolify self-hosting platform. Use when deploying to Coolify, configuring build settings, setting environment variables, managing health checks, or performing rollbacks.

Conductor/subagent routing for tasks across multiple harnesses. Assesses task complexity and intent, then dispatches to the appropriate subagent model via a harness-specific adapter. Supports Anthropic (Claude), Cursor, OpenCode, Codex (OpenAI), and Pi adapters. Use via /dynamic-task command or direct invocation from other skills.

Spawn three parallel review subagents over the active transcript, surface learnings, and route each to a concrete edit on an existing skill. Use when the user says reflect.

Stanford STORM multi-perspective research method. Runs four sequential phases (multi-perspective scan, contradiction map, synthesis, peer review) in one thread to produce a PhD-level briefing on any topic. Use for "storm research X", "multi-perspective research", "STORM method", or any topic where a single-prompt answer is too shallow.

Review recently changed files for reuse, quality, and efficiency issues, then fix them. Use after completing a change, before declaring done.

Extend grill-me by producing or updating a CONTEXT.md or ADR-style decision record while stress-testing a plan. Use when the user wants to be grilled and wants the shared understanding persisted as a living document.

Build Gemini-backed agent workflows and adapters. Use for Google Gemini API/ADK-style runners, Gemini CLI workflow parity, tool calling, structured outputs, or adding a Gemini variant beside Cursor/OpenCode/Claude runners.

Generate an interactive PR review walkthrough as an HTML page. Fetches PR data via gh API, categorizes files into core vs mechanical changes, adds reviewer annotations, and renders diffs with moved-code detection. Use when the user pastes a GitHub PR URL and asks for a review, walkthrough, or summary, or says \"review this PR\".

Document solved problems with context, examples, and gotchas so learnings compound for the team. Use after completing a workflow worth remembering.

Convert vague requests into compact TCRO prompts with phase-specific constraints. Used by research, specify, plan, and work workflows.

Continuous verification after each change, plus claim-based proof when evidence is required. Use when implementing, fixing bugs, or refactoring.

Long-running Jira investigation that self-continues past natural stops until a written hypothesis with evidence citations is produced. Pairs with the cooking Stop hook and hard-stops at 10 iterations.

Use for "automate me", "create/update/refresh my -mode skill", "turn/capture my preferences or working style into a skill", or wanting agents to follow how the user works. Drafts or revises a personal -mode skill via create-skill + unslop, optionally pulling fresh evidence from recent transcripts.

poteto's agent style for concise, detailed responses, deliberate subagents, unslopped prose, simple code, and verified work. Use for poteto, /poteto-mode, or requests to work in this style.

Discovery and invocation of all skills, commands, and agents. Decision tree for task-to-skill mapping. Core operating behaviors for agents. Use when unsure which skill or agent to use.

Simplifies code for clarity without changing behavior. Use when code works but is harder to read, maintain, or extend than it should be.

Extract reusable patterns from sessions, score confidence, and promote high-confidence instincts into durable skills. Use after completing significant work.

Coordinate a multi-repo refactor by fanning out subagents per repo onto one feature branch. Use for cross-repo changes that touch multiple repositories.

Review changes, open PRs, run pre-launch checks, staged rollouts, and rollback. Use when preparing to ship to production or closing out a branch.

Classify every skill in skills/ as user-invoked or model-invoked, ensure disable-model-invocation is set on user-invoked skills, and inject metadata.category on all. Run after adding or changing skills, or when token-cost/routing quality is being reviewed. Use for "sync skill taxonomy", "classify skills", "fix disable-model-invocation", or "skill token audit".