Extract shared Runway context and reduce skill repetition

README.md

@@ -127,11 +127,10 @@ The skills generate framework-specific code for:
 
 ## API Reference
 
-- **Base URL:** `https://api.dev.runwayml.com`
-- **Auth header:** `Authorization: Bearer <RUNWAYML_API_SECRET>`
-- **Version header:** `X-Runway-Version: 2024-11-06`
+See [`runway.md`](runway.md) for authentication, request headers, base URLs, output URL expiry, task polling, and other shared platform rules.
+
 - **Official docs:** [docs.dev.runwayml.com](https://docs.dev.runwayml.com/)
-- **API Reference:** [docs.dev.runwayml.com/api](https://docs.dev.runwayml.com/api)
+- **API reference:** [docs.dev.runwayml.com/api](https://docs.dev.runwayml.com/api)
 - **Developer portal:** [dev.runwayml.com](https://dev.runwayml.com/)
 
 ## License

runway.md

@@ -0,0 +1,78 @@
+# Runway API — Shared Context
+
+Durable platform rules that apply across all Runway API skills. Individual skills reference this document for common guidance instead of restating it.
+
+## Authentication
+
+- **Environment variable:** `RUNWAYML_API_SECRET`
+- **Server-side only.** The API key must never appear in client-side code, browser bundles, or public repositories.
+- Obtain a key at [dev.runwayml.com/settings/api-keys](https://dev.runwayml.com/settings/api-keys).
+- For staging environments, also set `RUNWAYML_BASE_URL=https://api.dev-stage.runwayml.com`.
+
+## Request Headers
+
+All REST requests require:
+
+```
+Authorization: Bearer <RUNWAYML_API_SECRET>
+X-Runway-Version: 2024-11-06
+Content-Type: application/json
+```
+
+SDK clients (`@runwayml/sdk` for Node.js, `runwayml` for Python) set these automatically.
+
+## Base URLs
+
+| Environment | URL |
+|-------------|-----|
+| Production  | `https://api.dev.runwayml.com` |
+| Staging     | `https://api.dev-stage.runwayml.com` |
+
+## Output URLs and Storage
+
+- Generated output URLs (video, image, audio) **expire within 24–48 hours**.
+- `runway://` upload URIs expire after **24 hours**.
+- Always download outputs to your own storage (S3, GCS, local filesystem) immediately after generation.
+- Never serve signed output URLs directly to end users.
+
+## Task Polling
+
+Generation endpoints return a task ID. The task must be polled until it reaches a terminal status.
+
+**SDK (recommended):**
+```javascript
+const task = await client.textToVideo.create({ ... }).waitForTaskOutput();
+```
+
+`waitForTaskOutput()` has a default 10-minute timeout. For long-running generations, implement a custom polling loop or increase the timeout.
+
+**REST:**
+```
+GET /v1/tasks/{id}
+```
+
+Poll until `status` is `SUCCEEDED`, `FAILED`, or `CANCELLED`. Use 5-second intervals with exponential backoff.
+
+## Credits and Billing
+
+- A Runway developer account with **prepaid credits** is required (minimum $10).
+- Credit costs vary by model — see individual skill docs or the `api-reference` skill for current pricing.
+- Manage billing at [dev.runwayml.com](https://dev.runwayml.com/).
+
+## Rate Limits
+
+- Rate limits are per-organization and vary by tier and model.
+- Check current limits via `GET /v1/organization`.
+- On `429 Too Many Requests`, retry with exponential backoff.
+
+## Content Moderation
+
+- Both inputs and outputs are subject to content moderation.
+- Safety-flagged inputs are **non-refundable** — credits are still consumed.
+- If a generation is rejected, try a different prompt.
+
+## API Reference
+
+- **Official docs:** [docs.dev.runwayml.com](https://docs.dev.runwayml.com/)
+- **API reference:** [docs.dev.runwayml.com/api](https://docs.dev.runwayml.com/api/)
+- **Developer portal:** [dev.runwayml.com](https://dev.runwayml.com/)

skills/integrate-audio/SKILL.md

@@ -172,7 +172,8 @@ async def generate_sound(req: SoundRequest):
 
 ## Tips
 
-- **Output URLs expire in 24-48 hours.** Download audio files to your own storage.
+See `runway.md` in the repository root for shared platform rules (output URL expiry, task polling, credits). Model-specific tips:
+
 - **For local audio files** (voice isolation, dubbing, speech-to-speech), upload via `+integrate-uploads` first.
 - **Voice IDs** can be listed via the voices endpoint — see `+api-reference` for details.
 - **Text-to-speech cost** scales with text length: 1 credit per 50 characters.

skills/integrate-image/SKILL.md

@@ -209,7 +209,8 @@ async def generate_image(req: ImageRequest):
 
 ## Tips
 
-- **Output URLs expire in 24-48 hours.** Download images to your own storage immediately.
+See `runway.md` in the repository root for shared platform rules (output URL expiry, task polling, credits). Model-specific tips:
+
 - **Reference images use `@Tag` syntax** in the prompt — the tag must match the `tag` field in the `referenceImages` array.
 - **For local files**, always upload via `+integrate-uploads` first, then use the `runway://` URI.
 - **`gen4_image_turbo`** is the cheapest option at 2 credits per image — good for prototyping.

skills/integrate-video/SKILL.md

@@ -289,9 +289,9 @@ async def generate_video(req: VideoRequest):
 
 ## Tips
 
-- **Output URLs expire in 24-48 hours.** Download videos to your own storage (S3, GCS, local filesystem) immediately after generation.
+See `runway.md` in the repository root for shared platform rules (output URL expiry, task polling, credits). Model-specific tips:
+
 - **`gen4_turbo` requires an image** — it cannot do text-only generation.
 - **`gen4_aleph` is the only video-to-video model** — use it for editing/transforming existing videos.
 - **Duration range is 2-10 seconds.** Longer videos require chaining multiple generations.
-- **`waitForTaskOutput()` has a default 10-minute timeout.** For long-running generations, you may want to implement your own polling loop or increase the timeout.
 - **For local files**, always use `+integrate-uploads` to upload first, then pass the `runway://` URI.

skills/recipe-full-setup/SKILL.md

@@ -96,9 +96,6 @@ Does the input come from a public HTTPS URL?
 
 ## Important Reminders
 
-- **Never expose the API key in client-side code.** All API calls must happen server-side.
-- **Output URLs expire.** Always download and store generated content.
-- **Credits are required.** The API won't work without prepaid credits.
-- **Rate limits exist.** Rate limits exist. You should always check what is the rate limit before attempting concurrent generations.
-- **Content moderation applies** to both inputs and outputs. Safety-flagged inputs are non-refundable.
+See `runway.md` in the repository root for shared platform rules (auth, output expiry, credits, rate limits, moderation). Key points for this workflow:
+
 - **Be cost-conscious.** Help users pick the right model for their budget. Credit cost can be found on https://docs.dev.runwayml.com/guides/pricing/