akios

Drives the data-modeling-canvas app — a React/Vite canvas for visually designing database schemas (entities, attributes, relations) — live in the browser via Chrome automation, instead of asking the user to paste JavaScript into DevTools. akios owns and runs this app itself: it installs once to ~/.akios/artifacts/data-modeling-canvas (source ships in this kit's artifacts/data-modeling-canvas/, shared across every project) and starts it on demand. Trigger when the user asks to model a database, design tables/entities and their relationships visually, wants a visual schema/ERD-style canvas, or mentions "Data Modeling Founder Lens" or the agent-canvas-plugin. Do NOT trigger for SQL migrations, SwiftData/Core Data model files, or any request that doesn't involve the live visual canvas app.

Whole-app mapping session — runs the FounderLens Double Diamond on the entire product, cartographs every major surface area (screens, data domains, flows, infrastructure), scopes each area (core/enhance/future), and bursts out a complete family of versioned specs into akios/specs/ and akios/Roadmap.md. Use when the user wants to map the full app before building, runs /akios:deep-brainstorm, says "map the whole app", "let's design everything", "generate all specs", "mapeie o app inteiro", or wants a comprehensive backlog before starting feature execution. Runs interactively by default; unattended deepthink mode when called under just-vibes.

Turns raw material (code, a PDF/book, an image/screenshot, .md docs, or a skill) into a knowledge pack the pipeline can route to. Use when the user wants akios to learn a new domain from material they trust, or runs /akios:learn. Routes extraction to the dedicated skill for each source type (pdf, pdf-reading, docx, multimodal, oss-first) and never hand-parses what a mature tool already does.

Turn an approved spec into a lean, executable backlog of task files under akios/tasks/todo/ in a single pass — the plan phase for Swift/iOS work. Use after idea-to-spec has produced akios/specs/<feature>.md and you need the execution backlog, or when a user runs /akios:plan. Produces atomic task files with est_tokens + runner, parallel markers, checkpoint barriers, definitions of done, and per-task UI-state coverage. Does NOT write app code — it writes the plan that task-execution runs.

Drive a akios/tasks/todo/ backlog to implemented, committed, reviewed Swift/iOS code — the kit's execution phase (absorbs the execution discipline once borrowed from superpowers). Use after spec-to-tasks has produced task files, or when a user runs /akios:deliver. Works on a per-spec branch, moves task files through folder states, writes tests first, commits at each checkpoint, manages the context window, archives finished specs, and stops at a hard human gate before any push or merge.

The design phase's one skill — explores multiple SwiftUI

Become FounderLens in the conversation itself — act as the virtual co-founder that walks a user's raw idea through the Double Diamond first diamond (Discover → Define) and a Midpoint Validation Audit, one ingredient at a time, ending in a versioned spec. Use when the user wants to run their idea through FounderLens directly in chat, be coached like a co-founder, or experience the pipeline without an interface. Speaks the user's language.

Master router for ALL Swift and iOS development. This skill classifies the scope of a change and loads the right specialized guides (bundled inside this skill) before any code is written. Trigger for: - Writing, editing, refactoring, or reviewing any .swift file or Xcode project - SwiftUI views, navigation, layout, modifiers, state management - Swift concurrency (async/await, actors, Sendable), SwiftData, Swift Testing - Debugging, running, or profiling an iOS/iPadOS/macOS/watchOS app - Implementing a Figma design in SwiftUI - Even when the user doesn't say "Swift" but the project or files are clearly Swift/Xcode-based Don't trigger for: - Other languages/platforms: Kotlin/Android, React Native, Flutter, web frontends, backend services not in Swift - Conceptual iOS questions with no code involved (App Store policy, app ideas, pricing) - Pure design work in Figma with no SwiftUI implementation requested

Pre-implementation UI alignment gate. Walks the user decision-by-decision through every visual and interaction choice for a screen before any SwiftUI code is written. Produces a ui-alignment doc that task-execution loads as implementation ground truth.

The agentic-kit meta-system for Swift/iOS repos — what it installs, how it routes work, and how to set it up. Use when setting up agentic Claude Code workflows in a Swift/iOS/iPadOS/macOS repo, when you see a CLAUDE.md that imports @AGENTS.md + @akios/Context.md, when deciding which gate/skill a Swift task routes to (idea-to-spec, oss-first, ios-feature-pipeline, spec-to-tasks, task-execution, swift-dev), or when installing/updating the kit.

Workflow orchestrator for taking a raw iOS feature idea all the way to working, reviewed code. Use whenever a user has a new feature idea for an iOS app and needs a structured path from concept to implementation. Triggers include: "I want to add X to the app", "let's build this feature", "new feature: ...", or any request to implement something non-trivial in a Swift/SwiftUI codebase. This skill does NOT write code itself — it reads akios/workflow.yml and routes you through the phases in order.

Collaborative design process for taking a raw product/system idea to versioned markdown specs through decision-by-decision co-design. Use whenever the user wants to design a complex product, tool, system, app, feature, or says things like "help me design X", "let's spec out this idea", "I want to turn this idea into a spec", "let's map the requirements", or wants a structured ideation partner that proposes options, waits for decisions, and produces spec documents.

Autonomous akios run. Drives the whole feature pipeline (brainstorm→plan→design→deliver→ship) end to end with no human in the loop — picks the next fuel (a submitted idea, the task backlog, designed-but-unbuilt specs, or akios/Vision.md/akios/Roadmap.md items), builds it, gates on quality, and ships it. Use when the user runs /akios:just-vibes (or --force), says "just vibe on it", "run autonomously", "drive the backlog yourself", "vai sozinho", or otherwise asks akios to make progress without supervising each step. Default stops after one unit at the spec boundary; --force loops until fuel is exhausted or interrupted.

Compact the current conversation into a handoff document so another agent session can continue the work — or return results to the originating session. Use when pivoting to a different concern mid-session, when context is growing unwieldy, or when a sub-task should run in isolation and report back.

Scaffolds a new akios skill or knowledge pack and self-registers it (install-skills.sh's SKILLS array, a commands/ wrapper, the version-bump trio) in one pass, so a capability is never half-installed. Trigger for: - The user wants to add a new skill/gate/router to this kit ("create a skill for X", "add a new gate") - The user runs /akios:new-skill - just-vibes decides, mid-run, that it needs a recurring capability encoded as a skill - The user wants to scaffold a new knowledge pack shell (before filling it with knowledge-ingest) Don't trigger for: - Filling an existing pack's content from code/PDF/image/docs — that's knowledge-ingest (/akios:learn) - Editing an already-registered skill's behavior — just edit its SKILL.md directly - Any Swift/iOS app code — this only authors kit-internal skills/packs, never app features

Before generating complex code, extensive documentation, or file transformations by hand, search for an established open-source tool that already solves the problem deterministically — cheaper, faster, and higher quality than token-by-token generation. Use this skill proactively whenever a task is a "commodity problem" that mature tooling likely solves: generating API docs from a Swagger/OpenAPI spec, converting file formats, linting/formatting, scaffolding projects, generating types from schemas, creating changelogs, diagrams from code, database migrations, parsing, minification, image processing, and similar. Also use it when the user asks "is there a tool/library/package for X", "what's the best open source option for X", or complains about token cost or AI rewriting things that tools already do.