What are Agent Skills?
Agent Skills are structured playbooks that encode the complete API call sequence, async completion pattern, error handling, and rate limit considerations for a common Framelane workflow. They are designed to be loaded by AI agents (Claude, GPT-4, Cursor Agent, etc.) to eliminate the need for the agent to figure out the multi-step API flow from scratch.Loading skills
Skills are static, machine-readable YAML playbooks — no MCP or account needed to read them. Fetch any docs page as raw markdown by appending.md to its URL, or install the
self-contained Framelane skill for the full agent
workflow. Once you hold a key, drive the actual calls through the authenticated action
MCP at https://mcp.framelane.io/mcp.
Before authoring anything, read GET https://api.framelane.io/v1/capabilities — no key
required. It is derived from the same enums and models the API validates against, so it
cannot drift: motions[] flags every preset with the surface it works on
(element_entrance, element_exit, character, text_block), effects /
transitions / blend_modes / easings list the full catalogs with a supported flag,
element_schemas gives the per-element JSON Schema, and request_schema gives the
composition level — groups, custom animations, motion blur, background gradient,
transitions.
MCP clients get the same document from the get_capabilities tool — call it with
no argument for a small index of every section and its size, then
section='request_schema' for the composition body and
section='element_schemas.text' (or .video, .image, .shape, .audio) for
one element type’s fields. On clients that also support resources the
same catalog is the framelane://capabilities resource; the tool exists because
resource support is not universal, so reach for the tool first. See the MCP
server page for the full tool list.
Available skills
Skill format
Each skill is a structured YAML document that encodes:- Goal: what the skill accomplishes
- Prerequisites: required API keys, webhook setup
- Steps: ordered API calls with exact request shapes
- Async pattern: how to wait for job completion
- Error handling: which error codes to expect and how to recover
- Rate limits: relevant limits to respect