agentleFS
GetStreamGetStream/agent-skills/skills/stream-android/SKILL.md

stream-android

Skill21 starsChanged 7 days agoApache-2.0View on GitHub

Build, integrate, and review Stream Chat, Video, and Feeds in Android apps. Use for Jetpack Compose, Android Studio, and Gradle project work - including Stream package setup, auth and token wiring, screen blueprints, read-only integration reviews before going live, AI assistants and LLM chat (ChatGPT-style apps, streaming replies, thinking indicators, stop generating) with stream-chat-android-ai-compose, and any follow-up Stream UI work such as adding screens, navigating between channel list and channel/message screens, channel tap handling, deep links, push routing, theming, custom channel/message UI, video calling flows (joining/starting calls, ringing, custom call controls and participant tiles), and Feeds surfaces (timeline, activity composer, threaded comments, follow graph / profile, notification feed, stories).

  • Reads credentials

1 other project uses a version of this, 1 of them edited.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 1

    1. nnawar706/fluento★ 094% same

    Wrote their own version · 0

    None found.

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository11 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

11 rules. About 0 are common, 9 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Rules: Read RULES.md once per session - every non-negotiable rule is stated there, nowhere else.

    line 28
  2. Distinctive:

    Before any tool call, decide the track from the user's input alone - no probes first.

    line 36
  3. Distinctive:

    If the request is ambiguous between build/integrate and reference lookup, ask one short question and wait:

    line 56
  4. Distinctive:

    run any product-specific setup (Chat: optionally seed channels.

    line 84
  5. Distinctive:

    Read-only local probe. Use it to detect whether the user is in an Android Studio / Gradle project, a Kotlin module, or an empty directory.

    line 92
  6. Distinctive:

    Use the result to produce a one-line status, for example:

    line 100
  7. Very specific:

    If the requested product/UI layer file is not bundled yet, say so plainly, use sdk.md for the shared Android patterns, and only switch to …

    line 146
  8. Distinctive:

    If the user asks for a product/UI-layer combo that is not bundled (e.g.

    line 192
  9. Distinctive:

    Use when the user wants the install and wiring path more than a feature build:

    line 198
  10. Distinctive:

    run Step 0.5 (credentials) first

    line 200
  11. Distinctive:

    stop before product-specific UI if the user only asked for setup

    line 205

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Stream Android - skill router + execution flow

Rules: Read RULES.md once per session - every non-negotiable rule is stated there, nowhere else.

This file is the single entrypoint: intent classification, local project detection, and module pointers for Stream work in Android apps.


Step 0: Intent classifier (mandatory first - never skip)

Before any tool call, decide the track from the user's input alone - no probes first.

Signals -> track

Signal in user inputTrack
Explicit product/framework token: Chat Compose, Chat XML, Video Android, Video Compose, Feeds Android, Feeds Compose, etc.C - Reference lookup
Words "docs" or "documentation" around Stream Android/Compose workC - Reference lookup
"How do I {X} in Compose/XML/Android?", "What does {SDK type/Composable/View/Fragment} do?"C - Reference lookup
"Build me a new Android app", "create a Compose app", "new Android app" + Stream productA - New app
"Add/integrate Stream into this app", "wire Chat/Video/Feeds into my Android project"B - Existing app
"Install Stream packages", "set up Stream in Android Studio", "wire auth/token flow" with no broader feature requestD - Bootstrap / setup
"Build a ChatGPT-style app", "add an AI assistant / support bot / copilot", "stream the LLM response", "thinking indicator", "stop generating", "voice input to the AI", LangChain, Vercel AI SDK, stream-chat-android-ai-composeA (new app) or B (existing app), then run ai-integration.md for the integration step; C if the user only asks how one AI component works
Request carries a target appearance - a screenshot, a Figma frame, or "make the chat look like WhatsApp / Telegram / Slack / <app>"B (or A) + the styling-depth flag below -> design-matching.md
"Migrate from Sendbird", "replace the Sendbird Chat SDK with Stream", "we're switching off Sendbird / SendbirdUIKit / uikit-compose"B + the migration flag below -> sendbird-migration.md
"Review / audit / check my Stream integration", "is this ready for production?", "prepare for the integration review"R - Integration review (read-only)
Bare /stream-android with no argsList the tracks briefly and wait

Disambiguation flow

If the request is ambiguous between build/integrate and reference lookup, ask one short question and wait:

Do you want me to wire this into the project, or just map the Android SDK pattern and files?

After classification

  • Tracks A, B, D -> run Step 0.5 (credentials) first, then Project signals once per session, then continue in builder.md and sdk.md. Do not probe the project before credentials.
  • Track C -> skip both steps if the product + UI layer are explicit. Only run Project signals on demand if the SDK or UI layer is ambiguous.
  • Track R -> skip credentials and Project signals (the review locates the integration itself); follow stream skill's integration-review.md with integration-review.md for the Android checks. Report first; change code only if asked.

Migration flag (orthogonal to the track)

If the user is migrating an existing app off Sendbird, run sendbird-migration.md — a repeatable Sendbird → Stream Chat procedure that sits on top of Track B. It detects the integration shape (Sendbird uikit-compose / XML uikit / custom UI on the low-level sendbird-chat SDK / spaghetti; MVVM/MVI; single- vs multi-activity; Compose-Nav vs fragments+Jetpack-Nav) and picks one of three paths: UIKit → Stream Compose bundled screens; custom-UI → keep the app's UI, swap only the SDK/state layer (surgical); XML UIKit → Stream Compose hosted in the existing fragments (cross-framework). It preserves the app's architecture and navigation, changing only SDK touchpoints. The load-bearing part is the gap-handling protocol (§3.5): for every screen the Sendbird source exposed that a naive swap would drop (creation, settings, members, invitation), either wire the bundled Stream equivalent (GroupChannelInfoScreen/DirectChannelInfoScreen) or — for screens Stream doesn't bundle (channel creation) — tell the user it isn't out-of-the-box, link the public cookbook (getstream.io/chat/docs/sdk/android/compose-cookbook/<slug>/), and offer to build it from that cookbook. Never silently drop a screen and never invent one. It reuses Step 0.5's credential flow (credentials.md) and the region rigor of design-matching.md for per-screen fidelity, and — once the code migration builds, connects, and matches the source — offers the data migration (the shared, language-agnostic runbook).

Styling-depth flag (orthogonal to the track)

If the request carries a target appearance - an attached screenshot, a Figma frame, or "make the chat look like WhatsApp / Telegram / Slack / <app>" - run design-matching.md before writing UI code. It sits on top of the Track A/B work (it is how you build the requested UI), not instead of it. A reference is a checklist of regions (header, composer, where the timestamp + read receipts sit, bubble shape/tail, channel-list rows, date separators, attachments...), and most differ from Stream's Compose defaults structurally, not just by color - setting the accent + wallpaper and stopping is the known failure. Classify each screenshot to its screen (channel list vs message vs thread - Step 0), decompose every region capturing dimensions not just colors, then route each difference to one of the two axes: a theming token (StreamDesign.Colors/Typography) or a ChatComponentFactory slot (structure - padding / corner-radius / bubble-shape / composer layout live here; Android has no Styles axis, so those are structure, not theming). Recurring traps the doc guards against: overriding a composite slot (MessageContainer, the header, MessageComposer) silently drops the sub-features the default drew (incoming avatar, grouping, reactions, replies, status, send/voice) unless you reproduce them; the message-header trailing slot defaults to a channel avatar (override it for action icons); the composer differs by row count / field shape / alignment far more than by icon set; DMs must not be prefixed with # in the list or titled "Channel" in the header. The match is not done until you build, seed data that triggers every region, compare region-by-region on the real navigation path, and iterate - implement every region, the composer included; a region left at the SDK default is a FAIL, not a footnote.


Step 0.5: Credentials, token, and seed data (tracks A, B, D only)

Order: intent classification -> Step 0.5 (this step) -> Project signals probe -> track work. Do this before running the Project signals shell command, even if a track table below lists "Detect" as phase 1. Run once per session for tracks A, B, and D. Skip for Track C.

Follow credentials.md to:

  • collect the Stream API key from the dashboard (or from the user)
  • generate a user token via the Stream CLI (or accept one from the user)
  • run any product-specific setup (Chat: optionally seed channels; Feeds: confirm feed groups; Video: nothing — calls are ephemeral)

Wire the API key from getstream env's output (e.g. BuildConfig.STREAM_API_KEY) and use the real user token in code snippets - never placeholder strings. If a track A/B/D task reaches code work and credentials haven't been collected yet, return to credentials.md before continuing.


Project signals (tracks A/B/D - once per session; Track C on demand only)

Read-only local probe. Use it to detect whether the user is in an Android Studio / Gradle project, a Kotlin module, or an empty directory.

bash -c 'echo "=== GRADLE ROOT ==="; find . -maxdepth 2 \( -name "settings.gradle.kts" -o -name "settings.gradle" -o -name "build.gradle.kts" -o -name "build.gradle" \) -print 2>/dev/null; echo "=== APP MODULES ==="; find . -maxdepth 3 -type f \( -name "build.gradle.kts" -o -name "build.gradle" \) -path "*/*/build.gradle*" -print 2>/dev/null; echo "=== VERSION CATALOG ==="; find . -maxdepth 3 -name "libs.versions.toml" -print 2>/dev/null; echo "=== MANIFESTS ==="; find . -maxdepth 4 -name "AndroidManifest.xml" -print 2>/dev/null; echo "=== EMPTY ==="; test -z "$(ls -A 2>/dev/null)" && echo "EMPTY_CWD" || echo "NON_EMPTY"'

Hold the result in conversation context. Don't re-run it unless the user changes directory or the project shape clearly changed.

Use the result to produce a one-line status, for example:

  • Compose app detected - app/build.gradle.kts - libs.versions.toml present - ready for Stream wiring
  • Multi-module Gradle project detected - preserve existing module layout
  • XML / View-based app detected - keep current UI layer unless the user asks to migrate
  • No Gradle project found - user needs to create the app in Android Studio first

Module map

TrackModule(s)
A - New appbuilder.md + sdk.md + relevant reference files
B - Existing appbuilder.md + sdk.md + relevant reference files
C - Reference lookupsdk.md + relevant reference files
D - Bootstrap / setupbuilder.md + sdk.md
R - Integration reviewstream skill's integration-review.md + integration-review.md

Reference layout

Live documentation is looked up through the llms.txt manifests in references/DOCS.md - the entrypoint for guides and surfaces this pack does not bundle. The AI runbook (LLM assistants, streaming replies, agents) lives in ai-integration.md, on top of the shared backend contract in ../stream/ai-backend-contract.md.

Shared Android/Kotlin patterns live in sdk.md. The curated procedure for reproducing a reference design with the pre-built Compose components (the region -> ChatComponentFactory-slot map and the build/verify loop) lives in design-matching.md - loaded via the styling-depth flag above, on top of whichever track applies.

Product and UI-layer specifics live under references/ using a flat naming scheme that can grow with the full Stream Android surface:

  • Reference: references/<PRODUCT>-<UI_LAYER>.md
  • Blueprints: references/<PRODUCT>-<UI_LAYER>-blueprints.md

Current extracted modules:

  • Chat + Compose: references/CHAT-COMPOSE.md + references/CHAT-COMPOSE-blueprints.md
  • Chat + XML: references/CHAT-XML.md + references/CHAT-XML-blueprints.md
  • Video + Compose: references/VIDEO-COMPOSE.md + references/VIDEO-COMPOSE-blueprints.md
  • Feeds + Compose: references/FEEDS-COMPOSE.md + references/FEEDS-COMPOSE-blueprints.md

Feeds has no pre-built UI components. FEEDS-COMPOSE.md covers the headless data SDK (FeedsClient, FeedState, ActivityState); FEEDS-COMPOSE-blueprints.md is custom Composable scaffolding driven by those state flows. Load both for any Feeds request.

Future Android product coverage should stay in this naming family instead of creating more top-level skills.

If the requested product/UI layer file is not bundled yet, say so plainly, use sdk.md for the shared Android patterns, and only switch to live docs if the user asks.


Track A - New app

Full detail: builder.md - use the new-project path.

PhaseNameWhat you do
A1DetectAfter Step 0.5 (credentials), run Project signals. If there is no Android app yet, tell the user to create one in Android Studio first.
A2Choose laneConfirm product(s) and UI layer: Compose, XML/Views, or mixed.
A3Install + wireFollow builder.md + sdk.md, then load only the needed product references.
A4VerifyConfirm Gradle sync, ChatClient lifetime, auth, and first rendered screen.

Track B - Existing app

Full detail: builder.md - use the existing-project path.

PhaseNameWhat you do
B1DetectAfter Step 0.5 (credentials), run Project signals and inspect the existing app structure before editing.
B2PreserveKeep the current UI layer, dependency strategy (version catalog vs inline), and navigation setup unless the user asks for a migration.
B3IntegrateUse sdk.md for shared wiring, then load only the needed product reference files.
B4VerifyConfirm the requested Stream flow builds and renders inside the existing app.

Track C - Reference lookup

Load only the relevant files for the requested product and UI layer.

  • Shared lifecycle / auth / state patterns -> sdk.md
  • Chat Compose setup and gotchas -> references/CHAT-COMPOSE.md
  • Chat Compose screen structure -> references/CHAT-COMPOSE-blueprints.md
  • Chat XML setup and gotchas -> references/CHAT-XML.md
  • Chat XML screen structure -> references/CHAT-XML-blueprints.md
  • Video Compose setup and gotchas -> references/VIDEO-COMPOSE.md
  • Video Compose call/screen structure -> references/VIDEO-COMPOSE-blueprints.md
  • Feeds Compose SDK patterns -> references/FEEDS-COMPOSE.md
  • Feeds Compose blueprints -> references/FEEDS-COMPOSE-blueprints.md
  • AI assistants, LLM chat, streaming replies -> ai-integration.md
  • Anything not bundled above -> the manifests in references/DOCS.md; fetch the page before answering

If the user asks for a product/UI-layer combo that is not bundled (e.g. Video XML, Feeds XML), say that clearly instead of inventing API details.


Track D - Bootstrap / setup

Use when the user wants the install and wiring path more than a feature build:

  • run Step 0.5 (credentials) first
  • detect the project shape
  • choose Compose vs XML ownership
  • install Stream packages with the project's existing dependency strategy (version catalog or inline)
  • wire auth and ChatClient lifetime via sdk.md
  • stop before product-specific UI if the user only asked for setup

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.