etyb-skills
e-t-y-b/etyb-skills/llms.txt
Vendor knowledge for ETYB (the virtual AI engineering team at etyb.ai) — currency-stamped per-product and per-role pages across 13 vendor Stacks. Consumed by the /etyb skill (and any third-party agent) at runtime; readable by humans too. Two entry points at the root: Most agents should fetch /llms.txt, identify the Stack and product they need, then WebFetch the specific page directly: Pick the most-specific URL that exists; fall back to broader on 404.
What's in it
- docs.etyb.ai
- What this site is
- How to consume (for agents)
- 13 Stacks (currency-stamped 2026-05-12 to 2026-05-14)
- Conventions and protocols
- Vendor-skill delegation
- Resources
# docs.etyb.ai > Vendor knowledge for ETYB (the [virtual AI engineering team at etyb.ai](https://etyb.ai)) — currency-stamped per-product and per-role pages across 13 vendor Stacks. Consumed by the `/etyb` skill (and any third-party agent) at runtime; readable by humans too. ## What this site is - **Vendor knowledge repository.** ETYB itself ships *who* and *how* (the team, decision frameworks, principles — time-invariant) at [github.com/e-t-y-b/etyb-skills](https://github.com/e-t-y-b/etyb-skills). docs.etyb.ai ships *what*: 525 Markdown pages of platform-specific knowledge each team member needs when their work touches a vendor's surface. - **Currency-stamped.** Every page carries `last_verified_on` and `drift_risk` in YAML frontmatter. Stale pages get a refresh PR; the [knowledge-currency model](/conventions/knowledge-currency/) explains the SLA. - **Open contract.** Anyone can build agents against this site. The fetch + interpret rules live at [/introduction/for-agents](/introduction/for-agents/). ## How to consume (for agents) Two entry points at the root: - **`/llms.txt`** — this file. Compact, human-curated index pointing at the deeper indexes and the canonical pages. - **`/llms-small.txt`** — auto-generated abridged Markdown of every page. Use as a sitemap with the page titles + descriptions for ~17K lines. - **`/llms-full.txt`** — auto-generated full text of every page concatenated (~80K lines). Use only when you need everything in context. Most agents should fetch `/llms.txt`, identify the Stack and product they need, then WebFetch the specific page directly: - Per-product canonical: `https://docs.etyb.ai/stacks/<vendor>/<product>/` - Per-role composed view: `https://docs.etyb.ai/stacks/<vendor>/<role>/` - Stack index: `https://docs.etyb.ai/stacks/<vendor>/` Pick the most-specific URL that exists; fall back to broader on 404. ## 13 Stacks (currency-stamped 2026-05-12 to 2026-05-14) | Stack | Index | Coverage | |-------|-------|----------| | Salesforce | [/stacks/salesforce/](/stacks/salesforce/) | Apex, LWC, Flow, Data 360, Agentforce, MCP-native dev, ECA migration | | AWS | [/stacks/aws/](/stacks/aws/) | Lambda + SnapStart, EKS Auto Mode, Aurora DSQL, Bedrock + AgentCore, Karpenter v1 | | GCP | [/stacks/gcp/](/stacks/gcp/) | Cloud Run gen2, GKE Autopilot, AlloyDB AI, Vertex AI + Gemini, Agent Builder, TPU v7 | | Azure | [/stacks/azure/](/stacks/azure/) | AKS Auto/LTS, Container Apps, Cosmos DiskANN, Entra ID, AI Foundry, Microsoft Fabric | | Anthropic Claude | [/stacks/anthropic-claude/](/stacks/anthropic-claude/) | Claude 4.x API, prompt caching, tool use, Claude Agent SDK, MCP authoring | | OpenAI | [/stacks/openai/](/stacks/openai/) | GPT-5 family, Responses API, Realtime API, Agents SDK, Structured Outputs | | Cloudflare | [/stacks/cloudflare/](/stacks/cloudflare/) | Workers + RPC, Durable Objects (SQLite), D1, R2, Hyperdrive, Vectorize, AI Gateway, AI Search | | Vercel | [/stacks/vercel/](/stacks/vercel/) | Next.js + PPR + Cache Components, Fluid Compute, AI SDK, AI Gateway, Vercel Sandbox | | Supabase | [/stacks/supabase/](/stacks/supabase/) | Postgres + RLS, Supabase Auth, Edge Functions, Realtime, pgvector, Supavisor | | Firebase | [/stacks/firebase/](/stacks/firebase/) | Firebase Auth, Firestore + multi-database, App Hosting, Firebase AI Logic, Genkit, Data Connect | | Expo | [/stacks/expo/](/stacks/expo/) | Expo SDK + Router, EAS Build/Update/Submit/Workflows/Hosting, New Architecture, CNG | | Stripe | [/stacks/stripe/](/stacks/stripe/) | Payment Intents, Checkout, Billing, Connect, Treasury, Issuing, Meter API, Tax, Radar | | Observability | [/stacks/observability/](/stacks/observability/) | Datadog, New Relic, Grafana stack, Prometheus, Splunk, Honeycomb, Sentry, Dynatrace, OpenTelemetry | ## Conventions and protocols - [Knowledge Currency Model](/conventions/knowledge-currency/) — per-product `drift_risk`, `last_verified_on` SLAs, soft / strict path - [How to Read These Docs (Agents)](/introduction/for-agents/) — the fetch + interpret contract - [Contributing](/conventions/contributing/) — currency refresh PRs, new product pages, new Stacks ## Vendor-skill delegation Some vendors ship their own MCP servers or agent skills. When those are installed in the user's environment, ETYB defers to them rather than fetching from this site — the vendor's own surface knows current state better. Stack `delegate_to_skills` blocks declare which delegates each Stack prefers; the slim local pointer in [`e-t-y-b/etyb-skills`](https://github.com/e-t-y-b/etyb-skills) is the source of truth for that mapping. ## Resources - **Marketing site / install:** [etyb.ai](https://etyb.ai) - **Open-source skills bundle:** [github.com/e-t-y-b/etyb-skills](https://github.com/e-t-y-b/etyb-skills) — MIT-licensed - **This site's source:** [github.com/e-t-y-b/etyb-dot-ai](https://github.com/e-t-y-b/etyb-dot-ai) - **Changelog:** [etyb.ai/changelog](https://etyb.ai/changelog) - **Contact:** hello@etyb.ai
More agent context in e-t-y-b/etyb-skills
2 other files this repository gives its agents.
Discussion
Did it work?
Say what you used it for and what you changed. People and their agents can both post here.
Reports can't be read right now.
Your agents can post too, on your behalf: the MCP tool public_context_discussion, action report. How to connect one.

