Update testing skills with marketplace plugin E2E testing knowledge

.agents/skills/testing-convert-stream-views/SKILL.md

@@ -0,0 +1,108 @@
+<!--
+SPDX-FileCopyrightText: © 2025 StreamKit Contributors
+
+SPDX-License-Identifier: MPL-2.0
+-->
+
+# Testing Convert & Stream Views End-to-End
+
+## Overview
+The Convert view (`/convert`) handles oneshot pipelines (file conversion, TTS, transcription) and the Stream view (`/stream`) handles dynamic MoQ pipelines. Both derive their UI controls from the declarative `client` section in pipeline YAML files.
+
+## Setup
+1. Start backend: `SK_SERVER__MOQ_GATEWAY_URL=http://127.0.0.1:4545/moq SK_SERVER__ADDRESS=127.0.0.1:4545 just skit`
+2. Start UI: `just ui`
+3. Navigate to `http://localhost:3045/convert` or `http://localhost:3045/stream`
+
+## Installing Marketplace Plugins for E2E Testing
+To test actual pipeline execution (not just UI rendering), install plugins via the marketplace:
+
+1. Enable marketplace in `skit.toml`:
+   ```toml
+   [plugins]
+   directory = ".plugins"
+   marketplace_enabled = true
+   allow_native_marketplace = true
+   registries = ["https://streamkit.dev/registry/index.json"]
+   models_dir = "models"
+   trusted_pubkeys = [
+       "untrusted comment: minisign public key 81C485A94492F33F\nRWQ/85JEqYXEgX+2kl7Rwd8AcpVjYciSLzvLggzivbGyIrDPjfmcqjYP\n",
+   ]
+
+   [plugins.security]
+   allow_model_urls = false
+   marketplace_scheme_policy = "https_only"
+   marketplace_host_policy = "any"
+   ```
+
+2. Install plugins via API:
+   ```bash
+   # Install Kokoro TTS
+   curl -X POST http://127.0.0.1:4545/api/v0/marketplace/install \
+     -H 'Content-Type: application/json' \
+     -d '{"id": "kokoro", "version": "0.1.0"}'
+
+   # Install Whisper STT
+   curl -X POST http://127.0.0.1:4545/api/v0/marketplace/install \
+     -H 'Content-Type: application/json' \
+     -d '{"id": "whisper", "version": "0.1.1"}'
+   ```
+
+3. Plugin bundles download to `.plugins/bundles/<name>/<version>/`
+4. Models download to `models/` directory (can be ~365MB for Kokoro, ~50MB for Whisper)
+5. After installation, restart the backend to load the new plugins
+
+## Convert View Pipeline Types
+
+### Text-to-Speech (client.input.type: text, client.output.type: audio)
+- Template: "Text-to-Speech (Kokoro)"
+- Shows textarea with character counter, no file upload
+- Button text: "Convert to Speech"
+- Section title: "Enter Text to Convert to Speech"
+- Output: Audio player with playback controls
+- Status shows "Streaming complete!" toast and "Completed (XX KB)"
+
+### Speech-to-Text (client.input.type: file_upload, client.output.type: transcription)
+- Template: "Speech-to-Text (Whisper)"
+- Shows "Upload File" and "Select Existing Asset" buttons
+- Asset list filtered by `asset_tags: [speech]` — only shows speech-tagged samples
+- Button text: "Transcribe Audio"
+- "Choose Output Mode" section is hidden (transcription mode)
+- Output: Transcription text with language badge, no audio player
+
+### No-Input (client.input.type: none, client.output.type: video)
+- Template: "Video Compositor Demo"
+- Input section is completely hidden
+- Button text: "Generate"
+- Output mode section renumbers to "3" (since input section is skipped)
+- Output: Video element with rendered content
+
+### Trigger (client.input.type: trigger)
+- Template: "Useless Facts TTS"
+- Similar to no-input but semantically represents a triggered action
+
+## Stream View Pipeline Types
+
+### Audio Publish+Watch (e.g. MoQ Peer Transcoder, MoQ Mixing)
+- Shows gateway URL (e.g. `/moq`, `/moq/mixing`)
+- Shows input broadcast name and output broadcast name
+- Audio-only: no video controls
+
+### Video Watch-Only (e.g. Video Color Bars MoQ)
+- Shows gateway URL (e.g. `/moq/video`)
+- Shows output broadcast name only (no input)
+- Video controls visible
+
+## Testing Tips
+- Use Playwright with headless Chromium for automated E2E testing
+- Pipeline conversions can be fast (<2s for short TTS) — use polling loops instead of fixed timeouts for completion checks
+- Wait for "Streaming complete" or "Transcription complete" toast notifications as completion signals
+- The Convert button re-enables after completion — this is another reliable completion indicator
+- For STT pipelines, use sample speech assets (speech_2m, speech_10m) rather than uploading files
+- Screenshot evidence at each step is valuable for verification
+
+## Relevant Files
+- `ui/src/views/ConvertView.tsx` — Main convert view component
+- `ui/src/views/StreamView.tsx` — Main stream view component
+- `samples/pipelines/oneshot/` — Oneshot pipeline YAML templates
+- `samples/pipelines/dynamic/` — Dynamic pipeline YAML templates