Agent Skills

Install the Tweetsmash skill so your coding agent (Claude Code, Cursor, and friends) knows how to work with your X (Twitter) bookmarks — search, read, organize, run smart views, share, export, and more — without you spelling out the API each time.

Install

Pick whichever fits your agent. All three install the same skill from tweetsmash/agent-skills.

skills CLI

The quickest way, for any agent that supports the open skills format.

Claude Code plugin

Install as a first-class Claude Code plugin from the Tweetsmash marketplace.

Manual

Clone the repo and drop the skill folder into your agent's skills directory.

skills CLI

bash

npx skills add tweetsmash/agent-skills

Claude Code plugin

claude code

/plugin marketplace add tweetsmash/agent-skills
/plugin install tweetsmash@tweetsmash

Manual

bash

git clone https://github.com/tweetsmash/agent-skills
cp -r agent-skills/skills/tweetsmash ~/.claude/skills/

Connect your account

Create an API token under Developer API. The skill uses the hosted MCP server (https://mcp.tweetsmash.com/api/mcp?token=<token>) or the REST API (https://api.tweetsmash.com/v1) — see Connect your agent for per-client config.

A bundled zero-dependency helper (Node 18+) can store and verify your token: node skills/tweetsmash/scripts/tweetsmash.mjs setup <token>.

What it teaches

The skill gives the agent the full Tweetsmash tool catalog and the rules to use it well — read a saved thread or X Article in full, bulk-tag or archive by filter, run and read Smart Views, create share links, export to PDF/CSV/JSON, manage email digests, and surface the top tweets from a Twitter list. It also knows the plan limits, so it prompts you to upgrade instead of failing silently.

The skill definition

This is the exact SKILL.md the agent reads:

Show SKILL.md

SKILL.md

---
name: tweetsmash
description: Work with a user's X (Twitter) bookmarks on Tweetsmash — read full content (articles, threads, notes), search, organize and label, run smart views, share, export, manage email digests, and discover high-signal tweets from their Twitter lists.
---

# Tweetsmash agent skill

Tweetsmash is where a user saves, organizes, reads, and searches their X (Twitter)
bookmarks. Use it to find and read saved tweets (full X Article bodies and unrolled
threads), organize them, share or export them, run their Smart Views, manage email
digests, and surface high-signal tweets from the Twitter lists they follow.

## Setup
- Get an API token: https://www.tweetsmash.com/integrations/api
- Preferred: connect the hosted MCP server — https://mcp.tweetsmash.com/api/mcp?token=<token>
- Or call REST at https://api.tweetsmash.com/v1 with Authorization: Bearer <token>
- Free plan: reading works but is limited (no pagination, small cap, no filters or
  semantic search). Filters, pagination, semantic search, exports and writes need a
  paid plan (HTTP 402 with an upgrade_url).

## Tools
- Bookmarks: list_bookmarks, search_bookmarks, get_bookmark (FULL content via
  include=article,thread,notes,replies), get_bookmark_count, setup_vector_store
- Organize: list_labels, add_labels_to_tweets, remove_labels_from_tweets,
  tag_bookmarks_by_filter, archive_bookmarks, archive_bookmarks_by_filter
- Smart Views: list_smart_views, get_smart_view, get_smart_view_bookmarks,
  create_smart_view, update_smart_view, delete_smart_view
- Share & export: create_share_link, revoke_share_link, trigger_pdf_export,
  trigger_csv_export, trigger_json_export
- Digests: list_digest_preferences, update_digest_preference,
  delete_digest_preference, list_past_digests
- Twitter Lists: list_twitter_lists, get_list_tweets (sort=LATEST|POPULARITY),
  get_list_digests, send_list_digest
- Sync: get_sync_status, get_bookmark_sync_status
- Account & memory: get_me, remember, recall, forget
- Product help: search_docs

## Behavior rules
- Always send the token; never log/echo/store it. 401 = bad token; 402 = paid plan.
- Prefer bulk-by-filter over paginate-then-loop. Confirm before deleting.
- Read threads/articles with get_bookmark. Exports/digests return a job/confirmation,
  not a file — tell the user to check their Export Center or inbox.

Install

Pick whichever fits your agent. All three install the same skill from tweetsmash/agent-skills.

skills CLI

The quickest way, for any agent that supports the open skills format.

Claude Code plugin

Install as a first-class Claude Code plugin from the Tweetsmash marketplace.

Manual

Clone the repo and drop the skill folder into your agent's skills directory.

skills CLI

bash

npx skills add tweetsmash/agent-skills

Claude Code plugin

claude code

/plugin marketplace add tweetsmash/agent-skills
/plugin install tweetsmash@tweetsmash

Manual

bash

git clone https://github.com/tweetsmash/agent-skills
cp -r agent-skills/skills/tweetsmash ~/.claude/skills/

Connect your account

A bundled zero-dependency helper (Node 18+) can store and verify your token: node skills/tweetsmash/scripts/tweetsmash.mjs setup <token>.

What it teaches

The skill definition

This is the exact SKILL.md the agent reads:

Show SKILL.md

SKILL.md

---
name: tweetsmash
description: Work with a user's X (Twitter) bookmarks on Tweetsmash — read full content (articles, threads, notes), search, organize and label, run smart views, share, export, manage email digests, and discover high-signal tweets from their Twitter lists.
---

# Tweetsmash agent skill

Tweetsmash is where a user saves, organizes, reads, and searches their X (Twitter)
bookmarks. Use it to find and read saved tweets (full X Article bodies and unrolled
threads), organize them, share or export them, run their Smart Views, manage email
digests, and surface high-signal tweets from the Twitter lists they follow.

## Setup
- Get an API token: https://www.tweetsmash.com/integrations/api
- Preferred: connect the hosted MCP server — https://mcp.tweetsmash.com/api/mcp?token=<token>
- Or call REST at https://api.tweetsmash.com/v1 with Authorization: Bearer <token>
- Free plan: reading works but is limited (no pagination, small cap, no filters or
  semantic search). Filters, pagination, semantic search, exports and writes need a
  paid plan (HTTP 402 with an upgrade_url).

## Tools
- Bookmarks: list_bookmarks, search_bookmarks, get_bookmark (FULL content via
  include=article,thread,notes,replies), get_bookmark_count, setup_vector_store
- Organize: list_labels, add_labels_to_tweets, remove_labels_from_tweets,
  tag_bookmarks_by_filter, archive_bookmarks, archive_bookmarks_by_filter
- Smart Views: list_smart_views, get_smart_view, get_smart_view_bookmarks,
  create_smart_view, update_smart_view, delete_smart_view
- Share & export: create_share_link, revoke_share_link, trigger_pdf_export,
  trigger_csv_export, trigger_json_export
- Digests: list_digest_preferences, update_digest_preference,
  delete_digest_preference, list_past_digests
- Twitter Lists: list_twitter_lists, get_list_tweets (sort=LATEST|POPULARITY),
  get_list_digests, send_list_digest
- Sync: get_sync_status, get_bookmark_sync_status
- Account & memory: get_me, remember, recall, forget
- Product help: search_docs

## Behavior rules
- Always send the token; never log/echo/store it. 401 = bad token; 402 = paid plan.
- Prefer bulk-by-filter over paginate-then-loop. Confirm before deleting.
- Read threads/articles with get_bookmark. Exports/digests return a job/confirmation,
  not a file — tell the user to check their Export Center or inbox.