agentsclimarketplace

Android architecture design desk

Skill MadewellRD/skills-lab/dist/vendor/google/android-command-desk/android-architecture-design-desk

design Android app and game architecture, module boundaries, data flow, offline behavior, engine integration, services, APIs, migrations, and ADR-ready decisions.From its SKILL.md

Install
npx -y skills add MadewellRD/skills-lab --skill android-architecture-design-desk

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

7.5 KB, ~1.5k tokens by cl100k_base, as published. Nobody here has run it

Android Architecture Design Desk

Suite workflow mode

This desk is part of the Android Command Desk workflow suite. Complete this desk's artifact, update the android_delivery_packet, and continue when enough source facts are available. Return Workflow Halt instead of inventing repo architecture, runtime constraints, service contracts, migration scope, or validation facts.

Role

Define Android architecture for app and game work: modules, layers, state, data flow, storage, networking, engine boundaries, native/plugin boundaries, background work, API contracts, migration impact, ADR decisions, and validation gates.

Workflow

Outcome. An Android architecture decision set ready for ADR capture: architecture lane, module boundaries and ownership, data and state flow, threading and concurrency, storage, networking, background work, service contracts, engine/native/plugin boundaries, migration impact, rejected alternatives, risks, and validation gates.

Grounding. Build on accepted requirements, technical discovery facts, and the repo's current structure. Do not invent repo architecture, module ownership, runtime constraints, service contracts, migration scope, or validation facts.

Ordered content that stays ordered. Where the design includes an on-device data migration, a schema change, a storage-engine swap, or a save-format change, emit its steps as an ordered sequence and keep them ordered. The order is mandated by the device, not by this desk: a migration applied out of order against user data on a shipped install is irreversible and cannot be undone by a subsequent release. Do not collapse those steps into prose.

Parallel surface. Candidate module boundaries, individual service contracts, and per-alternative trade-off analysis are independent and can be evaluated in parallel. Lane selection, the module graph, and the ADR set are aggregate: they reconcile the parallel results and run once.

Acceptance bar. The architecture is done when the lane is chosen with rejected alternatives and their reasons recorded; every module boundary names its owner or marks ownership unknown; each decision is tied to a requirement or a discovery fact rather than to general Android practice; migration impact is stated or explicitly ruled out; and every validation gate is tied to source evidence.

Continue to UI/UX, app engineering, or game engineering based on target outcome.

Responsibilities

  • Ground architecture in current repo structure and Android app/game evidence.
  • For app work, account for Kotlin, Compose/View, modularization, DI, data layer, offline/sync, Macrobenchmark, and Baseline Profile implications.
  • For game work, account for engine/runtime boundaries, native libraries, asset delivery, frame loop, input, and profiling constraints.
  • Produce ADR-ready decisions before implementation handoff.

Expected inputs

Android PRD, discovery memo, repo facts, existing architecture docs, service contracts, design/game docs, performance/security constraints, and prior android_delivery_packet.

Expected outputs

A complete run produces the full design package in one pass: the architecture brief, the ADR notes, the module and interface map, the migration plan, the risks, the validation expectations, and the packet update. They interlock; an ADR without the module map it constrains, or a migration plan without the risks that gate it, is not something an engineer can build against, so treat them as one deliverable rather than a list to choose from.

Each artifact is done when implementation could start from it. The brief names the selected lane and the alternatives that lost and why; the module map names real modules, real ownership, and the interfaces between them rather than boxes labelled by layer; the migration plan states the sequence, the reversible points, and what breaks if a step is skipped. Section headings with nothing decided underneath are a failed artifact, not a first draft.

Producing everything is not licence to design past the evidence. Where no source establishes a module boundary, a dependency direction, a concurrency constraint, or a migration dependency, mark that part blocked on the named missing fact; a convincing architecture invented to fill the gap is worse than an acknowledged hole. Independent per-boundary and per-contract work is part of the parallel surface declared in Workflow.

Evidence packet additions

  • architecture lane and rejected alternatives
  • module boundaries and ownership
  • data/state flow, storage, networking, and background work
  • engine/native/plugin boundaries for game work
  • ADR decisions, risks, and validation gates

Packet fields to update

architecture_lane, modules, interfaces, data_flow, state_management, storage, networking, background_work, engine_runtime, migrations, risks, validation_commands, source_facts, open_questions, ready_to_continue

Halt conditions

Proceed by default. An unresolved design detail is normally a labeled assumption plus an open question, not a stop. Reserve hard halts for these consequence classes:

  • Approval: the architecture commits to a platform, vendor, or cost decision that a human owner must authorize.
  • Production or destructive: the design requires a data migration, storage swap, or schema change against live user data with no approved rollback path.
  • Security or privacy: a boundary decision determines how personal data, credentials, or keys are stored or transmitted and no source establishes the requirement.
  • Source conflict: repo structure, existing architecture docs, and stated requirements genuinely disagree on a load-bearing boundary. Preserve the conflict.
  • Release integrity: an ADR would record a decision as accepted when no source shows it was accepted, or validation gates cannot be tied to any evidence.
  • Connector unreachable: repo or architecture-doc access exists but cannot be read.

Otherwise proceed: an unknown API contract, runtime constraint, architecture lane, or migration scope becomes a labeled assumption in the ADR plus an open question routed to the desk that can resolve it.

Default output modes

A complete run writes all of these:

  • android-architecture-brief.md
  • android-adr.md
  • android-module-map.md
  • android-migration-plan.md

Mode-specific alternative:

  • workflow-halt.md: stands in place of the set above when a hard halt fires, rather than accompanying a finished design. Where no migration is in scope, say so in the brief instead of writing an empty migration plan.

Where a file has no evidential basis, it names what is missing. A design document is never padded out to fill its slot in the list.

Downstream handoff

Continue to android-ui-ux-desk for screen/interaction work, android-app-engineering-desk for native implementation, or android-game-engineering-desk for game runtime work.

SDLC suite handoff

Use architecture-design-desk patterns for ADRs, interface contracts, migration planning, and architectural risks while preserving Android-specific platform and runtime constraints.

Capability baseline

Use references/capability-baseline.md for what may be assumed about the executing model: context budget, native self-verification, long-horizon continuation, and parallel fan-out. It also states the governance invariants that do not relax as models improve.

What ships with it: 2 files

5.7 KB alongside SKILL.md

agents/

references/

Gives 0 of the 12 instructions most design frontend skills give in ~1.5k tokens

Counted across 1,169 of the 1,878 authors here whose files we hold, read 2026-08-07

  • Use CSS variables for color consistencyin 72 of 1169, across 23 files
  • Commit to one bold aesthetic direction before codingin 72 of 1169, across 27 files
  • Match implementation complexity to the aesthetic visionin 70 of 1169, across 20 files
  • Add atmospheric background effects and texturesin 57 of 1169, across 9 files
  • Use unexpected spatial compositions and layoutsin 56 of 1169, across 8 files
  • Implement real working codein 55 of 1169, across 7 files
  • Vary themes and aesthetics across different designsin 48 of 1169, across 7 files
  • Launch chromium in headless modein 47 of 1169, across 4 files
  • Close the browser when donein 47 of 1169, across 4 files
  • Run provided scripts with help flag firstin 47 of 1169, across 4 files
  • Wait for network idle statein 47 of 1169, across 4 files
  • Use descriptive selectors for elementsin 47 of 1169, across 4 files

Said here and by no other author read

  • ground decisions in repo facts and evidence
  • produce ADR-ready decisions for implementation handoff
  • tie each decision to a specific requirement or discovery fact
  • emit data migration steps as an ordered sequence
  • do not collapse ordered migration steps into prose
  • name a real owner for every module boundary

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 326,764. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.