SA

selat-ai/selat-skills

开发工具

本页由 AI 翻译,技术细节请以原文为准。 查看原文

2 stars 质量 63 趋势 28Apache-2.0

Skill definitions for the SELAT agent-payments ecosystem. Each skill composes one or more catalogue API endpoints into a named capability, paid via selat-pay and the SELAT Router.

概览

Skill definitions for the SELAT agent-payments ecosystem. Each skill composes one or more catalogue API endpoints into a named capability, paid via selat-pay and the SELAT Router. Skills follow the authoring standard — see references/agent-skill-authoring-sop.md. This repo holds skill only; the CLI that lists, installs, and runs these skills lives in selat-cli (selat skill list|install|run). - makes the skill activatable and documented per the SOP (frontmatter name/description/license/compatibility/metadata, plus When To Use, Workflow, Inputs And Outputs, Gotchas, Validation, References). - is the inert, machine-readable recipe selat-cli executes (no code — just steps mapped to selat-pay calls). It is the one skill file the CLI fetches on selat skill install . - holds trigger and output assertions per the SOP. - — Circle nanopayment / Gateway-batched, paid straight to the upstream (no router hop).

README

以下为原文 README 节选,完整内容请查看原文。

selat-skills

Skill definitions for the SELAT agent-payments ecosystem. Each skill composes one or more catalogue API endpoints into a named capability, paid via selat-pay and the SELAT Router.

Skills follow the Agent Skill authoring standard — see references/agent-skill-authoring-sop.md. This repo holds skill content only; the CLI that lists, installs, and runs these skills lives in selat-cli (selat skill list|install|run).

Skill layout

Each skill is an Agent Skill directory:

skills//
├── SKILL.md            # required — frontmatter + operational docs (the SOP)
├── manifest.json       # machine-readable payment recipe (read by selat-cli)
└── evals/
    └── evals.json      # trigger + output-quality evals
  • SKILL.md makes the skill activatable and documented per the SOP (frontmatter name/description/license/compatibility/metadata, plus When To Use, Workflow, Inputs And Outputs, Gotchas, Validation, References).
  • manifest.json is the inert, machine-readable recipe selat-cli executes (no code — just steps mapped to selat-pay calls). It is the one skill file the CLI fetches on selat skill install .
  • evals/evals.json holds trigger and output assertions per the SOP.

Rails

  • direct — Circle nanopayment / Gateway-batched, paid straight to the upstream (no router hop).
  • routed — erc-3009 or tempo-native MPP, paid via the SELAT Router, which translates the agent’s inbound Gateway-batched payment to the upstream’s scheme.
  • mixed — a multi-step skill whose steps settle on more than one rail in one run (e.g. stock-direction-signals, which mixes x402 and MPP steps).

In the manifests, each step’s rail is a descriptive label read from the live probe (x402 via Circle Gateway, x402 on Base, MPP on Tempo, MPP on Solana). A skill whose steps all share one label lists that label in the table below; otherwise it lists mixed. A label names the merchant’s rail, not the chain you pay from: the CLI settles from your funded Circle Gateway balance.

Current coverage

This catalog’s vetted skills are listed below. Coverage is concentrated in two families:

  • B2B / GTM enrichment — lead, person, company, email, prospecting, funding, creator, and VC sourcing workflows.
  • Financial / social / web research — market and wallet intelligence, stock-direction research, Twitter/X research, social listening, Perplexity web search, and entity reputation briefs.

The table below and index.json are generated from each skill’s manifest.json by npm run catalog; CI fails if either is out of date.

Skills

20 skills. Generated from each skill’s manifest.json (rail and kind are derived from its steps).

在 GitHub 查看完整 README

推荐工具

换一个关键词,或者移除筛选条件。

安装

npx skillfish add selat-ai/selat-skills
新手?先看 mcpskill.org 教程 →