Skills and agent-readable resources

This page explains tools, skills, and documentation for users and agents. Reading instructions still requires actual setup, authentication, and free connection verification.

What each component does

Component	Purpose	Next step
EvoLink CLI	Executes platform operations in a terminal; media calls go directly to APIs	Install, sign in, and verify
Remote MCP	Exposes host tools over HTTP and OAuth	Choose a client
Local MCP	Runs tools in a local stdio process	Configure startup and a platform Key
Skill	Guides discovery, approval, recovery, and delivery	Install where the host discovers it and confirm loading
Plugin	Packages host-specific configuration and skills	Plugin steps, plus authentication and permission checks
Markdown/llms.txt	Readable instructions or a documentation index	Read linked pages; these do not execute tools

Installing CLI does not clone MCP/plugin repositories or register native MCP. CLI bundles evolink-cli; the MCP plugin repository provides evolink-mcp tool guidance. They remain separately maintained entry points, not an already unified skill source.

What the skills contain

Current skills cover models/parameters, image/video/audio workflows, references, estimates/approval, budget checks, original submission recovery, waiting, downloading, and original delivery. CLI includes SKILL.md and three references: capabilities, media-workflows, and usage-and-budget. See CLI skills for targets, directories, loading, and upgrades; see MCP plugins for host setup. There is no separate viral-ad, finished-UGC, automatic-editing, or preset skill library. Examples do not guarantee popularity. Code development, arbitrary file editing, and website building belong to the host rather than EvoLink tools.

Shared agent instructions

Choose EvoLink CLI or native MCP for this task and verify the actual connection.
Query account-available models and real schemas; explain selection by task, references, quality, and budget.
Show final specifications and the estimated cost, then submit once after explicit approval.
Recover unknown outcomes through the original route, identity, and request ID.
Query the original task_id and deliver originals, reported actual costs, and expiry.
Explain unsupported capabilities; do not invent models, inputs, tools, prices, or refunds.
Do not describe estimate budgets as final charge guarantees or automatically edit/regenerate results.

Bind the task to one route. CLI uses commands/local quotes; MCP uses tools/original client_request_id. Consult the appropriate reference. Installed skill files do not establish host loading, and reading skills does not configure tools or grant execution permission. Do not repeatedly generate to test discovery.

Use Markdown and llms.txt

Read the overview and choose the client/route.
Read its setup, command/tool reference, and shared rules. Query current model schemas during execution.
Use supported page copying or Markdown export to save a .md file for your agent. Keep Keys, tokens, and private authorization links out of it.
llms.txt indexes documentation; it is not an MCP endpoint, CLI installer, or authentication API.

Resource	Purpose
CLI reference / MCP reference	Exact operations, inputs, results, and recovery
Models, costs, and quotas	Selection, estimates, approval, and budget boundaries
Tasks, references, and results	Uploads, expiry, recovery, downloads, and gray previews
CLI skills / MCP plugins	Installation, discovery, personal edits, and upgrades

Verify after updates

Synchronize bundled skills after a CLI upgrade and reload them in the host. After plugin/manual MCP changes, check the current connection, OAuth, and permissions. Preserve other services and personal edits. Package, documentation, and tool versions can differ; resolve conflicts against actual released versions and returned schemas and explain uncertainty.

Start here

CLI

MCP

Shared resources

Skills and agent-readable resources

What each component does

What the skills contain

Shared agent instructions

Use Markdown and llms.txt

Verify after updates

​What each component does

​What the skills contain

​Shared agent instructions

​Use Markdown and llms.txt

​Verify after updates

What each component does

What the skills contain

Shared agent instructions

Use Markdown and llms.txt

Verify after updates