generated: '2026-08-13' method: searched source: https://github.com/svix/ai sources: - https://github.com/svix/ai - https://raw.githubusercontent.com/svix/ai/main/skills/README.md - https://raw.githubusercontent.com/svix/ai/main/plugin.json provenance: | PROVIDER-PUBLISHED, SAVED VERBATIM. These skills were not generated by this pipeline. Svix authors and maintains them in github.com/svix/ai (MIT) as the canonical source, and every file under skills/ in this repo is a byte-for-byte copy fetched from raw.githubusercontent.com on 2026-08-13. Nothing here was written on Svix's behalf. packaging: format: Agent Skills (https://agentskills.io/) plugin_spec: Agent Plugins v1.0.0 (https://agent-plugins.org/schemas/1.0.0/plugin.schema.json) plugin_manifest: https://github.com/svix/ai/blob/main/plugin.json plugin_name: svix plugin_version: 1.0.0 install: npx skills add svix/ai license: MIT structure: | Each skill is a SKILL.md carrying frontmatter (name, description, and in one case an allowed-tools allowlist) plus, for the larger skill, a references/ folder the agent pulls in on demand. Svix's own README describes SKILL.md as "a router: it holds the decision, not the content" — progressive disclosure so the agent does not spend context on Ingest guidance while working on Dispatch. skills: - file: svix-sending-webhooks/SKILL.md name: svix-sending-webhooks description: >- Everything for working with Svix webhooks: first-time setup (API key, SDK install, first message), Dispatch (sending webhooks to your customers), Ingest (receiving third-party webhooks), Applications, Channels, customer UIDs, idempotency, App Portal embedding, operational webhooks, the Svix CLI, and — only when the user explicitly asks for one — a written integration plan. api: openapi/_original/svix-openapi.json scope: Svix-specific method: searched upstream: https://github.com/svix/ai/blob/main/skills/svix-sending-webhooks/SKILL.md has_allowed_tools: true allowed_tools_note: | Declares a tool allowlist that pins WebFetch to docs.svix.com, api.svix.com, www.svix.com, svix.com, github.com and raw.githubusercontent.com, and restricts Read/Grep to the skill's own references and the working tree. Publishing a skill with an explicit domain allowlist is rare and worth noting. modes: - id: build default: true note: A request to integrate Svix is treated as a request for working code. - id: plan default: false note: Entered only on an explicit request for a plan; writes no code until confirmed. references: - file: svix-sending-webhooks/references/quickstart.md covers: First-time setup — API key, SDK install, first message. - file: svix-sending-webhooks/references/dispatch.md covers: Sending webhooks — tenancy, message.create, Channels, App Portal, operational webhooks. - file: svix-sending-webhooks/references/dispatch-questions.md covers: What to ask before designing a sending integration, including migration cutover. - file: svix-sending-webhooks/references/ingest.md covers: Receiving third-party webhooks — Source Types, fanout, transformations, handlers. - file: svix-sending-webhooks/references/ingest-questions.md covers: What to ask before designing an Ingest integration. - file: svix-sending-webhooks/references/cli.md covers: Svix CLI usage rules, `svix listen`, scripting with jq. - file: svix-sending-webhooks/references/planning.md covers: The plan-first workflow. - file: svix-sending-webhooks/references/triage.md covers: What to look for when investigating an existing repo. - file: svix-sending-webhooks/references/plan-template.md covers: The shape of the written integration plan. - file: receiving-webhooks/SKILL.md name: receiving-webhooks description: >- General guidelines for building a robust webhook receiver/handler: verifying signatures, raw-body access, replay protection, async processing, retries and endpoint auto-disabling. Use whenever you write, review, or debug a handler that consumes incoming webhooks from any provider. api: null scope: provider-agnostic method: searched upstream: https://github.com/svix/ai/blob/main/skills/receiving-webhooks/SKILL.md has_allowed_tools: false note: | Deliberately NOT Svix-specific. It teaches Standard Webhooks verification for any provider and only mentions Svix as a tell ("the svix-id, svix-timestamp and svix-signature headers"). A vendor publishing a vendor-neutral skill is an unusual move and the reason this one is worth keeping in the catalog: it is useful to an agent integrating with anybody. related_agent_surfaces: - kind: MCPServer url: https://docs.svix.com/ai/app-portal-mcp artifact: mcp/svix-mcp.yml - kind: agent-runtime-plugin name: svix-openclaw url: https://github.com/svix/ai/tree/main/svix-openclaw note: | Polls a Svix sink and hands each message to the OpenClaw runtime as if it were an inbound POST — webhook delivery into an agent with no public URL, no open port and no tunnel. - kind: agent-runtime-plugin name: svix-hermes url: https://github.com/svix/ai/tree/main/svix-hermes note: The same inversion for the Hermes Agent gateway. - kind: LLMsTxt url: https://docs.svix.com/llms.txt artifact: llms/svix-llms.txt note: | Every docs page also serves a markdown twin by appending `.md` to its path — documented in llms.txt itself and repeated inside the skill. not_generated: note: | This pipeline did NOT author any additional skills for Svix. The provider's own two cover the marquee flows the arazzo/ workflows already model (provision an application, send a message, recover failed deliveries, set up Ingest), and a generated skill would sit alongside a maintained first-party one asserting the same thing less well. gaps: - id: skills-not-in-docs-index detail: | plugin.json points at https://docs.svix.com/ai as the homepage and the repo README links https://docs.svix.com/ai/mcp. Both 404 as of 2026-08-13, and no /ai/ page appears in docs.svix.com/llms.txt. The skills are discoverable from GitHub and from an component embedded in the quickstart, but not from the documentation index — a real agent surface with a broken front door.