generated: '2026-08-11'
method: searched
source: https://github.com/tweetapi/agent-skill
note: >-
  Provider-published, not generated by API Evangelist. TweetAPI maintains an official Agent Skill
  repository (MIT, github.com/tweetapi/agent-skill) carrying a SKILL.md, three reference documents,
  a Claude Code plugin manifest, a Codex plugin manifest, an OpenAI agents descriptor and an
  .mcp.json client configuration. The SKILL.md and its references are saved here verbatim. Nothing
  in this directory was authored on the provider's behalf.
repository:
  url: https://github.com/tweetapi/agent-skill
  license: MIT
  default_branch: main
  last_pushed: '2026-08-07'
  topics: [agent-skills, ai-agents, claude-code, codex, cursor, mcp, model-context-protocol, oauth, tweetapi, twitter-api]
skills:
  - file: tweetapi-read-only-twitter-data.md
    name: tweetapi
    description: >-
      Integrate TweetAPI or use its hosted MCP server for bounded, read-only access to public
      Twitter/X data. Excludes posting, engagement, account access, private messages, login
      automation, billing and administration.
    api: https://api.tweetapi.com/tw-v2
    mcp: https://mcp.tweetapi.com/mcp
    source: https://github.com/tweetapi/agent-skill/blob/main/skills/tweetapi/SKILL.md
    method: searched
    operations: []
    operations_note: >-
      The skill deliberately names no operationIds. TweetAPI publishes no OpenAPI, and the skill
      instructs agents to read the current public sources rather than rely on a copied endpoint
      catalog ("Do not copy the complete endpoint catalog into this skill. It will drift.").
    references:
      - {file: references/tweetapi-hosted-mcp.md, source: 'https://github.com/tweetapi/agent-skill/blob/main/skills/tweetapi/references/hosted-mcp.md'}
      - {file: references/tweetapi-integration.md, source: 'https://github.com/tweetapi/agent-skill/blob/main/skills/tweetapi/references/integration.md'}
      - {file: references/tweetapi-sources.md, source: 'https://github.com/tweetapi/agent-skill/blob/main/skills/tweetapi/references/sources.md'}
client_manifests:
  - {kind: claude-code-plugin, path: .claude-plugin/plugin.json, url: 'https://github.com/tweetapi/agent-skill/blob/main/.claude-plugin/plugin.json'}
  - {kind: codex-plugin, path: .codex-plugin/plugin.json, url: 'https://github.com/tweetapi/agent-skill/blob/main/.codex-plugin/plugin.json'}
  - {kind: openai-agents, path: skills/tweetapi/agents/openai.yaml, url: 'https://github.com/tweetapi/agent-skill/blob/main/skills/tweetapi/agents/openai.yaml'}
  - {kind: mcp-client-config, path: .mcp.json, url: 'https://github.com/tweetapi/agent-skill/blob/main/.mcp.json'}
observations:
  - >-
    The skill draws a boundary narrower than the API. TweetAPI's REST surface includes posting,
    engagement, profile mutation, X Chat and legacy DMs; the skill forbids all of them and forbids
    the agent from ever handling authToken, ct0, TOTP codes, proxy credentials or message content.
  - >-
    It also imposes quota discipline an agent would not otherwise infer: no auto-pagination, no
    silent retry, no live smoke tests, one live call and one page by default, and an explicit
    statement to the user that a call consumes quota.
  - >-
    It instructs agents not to invent an OpenAPI URL, Postman collection, SDK method or MCP tool
    name — the provider's own acknowledgement that neither a public OpenAPI nor a Postman collection
    exists.