What does this skill do, and when should you use it?
This is the official Agent Skill from Typefully's repo, wiring an AI agent to the Typefully API across X (Twitter), LinkedIn, Threads, Bluesky, Mastodon, and Substack Notes. The core is a zero-dependency Node.js 18+ script whose commands all output JSON for reliable agent parsing. It handles draft creation, threads, media and tags, scheduling, planned (inert) drafts, queue timelines, publishing quotas, and dropped Typefully draft URLs. It is MIT-licensed and requires your own Typefully API key.
- Calls the Typefully API through ./scripts/typefully.js to create, update, list, and delete drafts
- Schedules or publishes with --schedule/--plan/--publish; next-free-slot lets Typefully pick the optimal time
- Publishes one draft to multiple platforms in a single call (--platform x,linkedin or --all); threads split on --- lines
- Uploads media, manages tags, and shows queue timelines and the publishing_quota
- Maps a pasted Typefully draft URL directly to a drafts:get command
- A creator who wants an AI agent in their IDE or terminal to handle daily X/LinkedIn posting and scheduling
- A team that needs one announcement posted to X and LinkedIn at once, or platform-tailored variants of the same content
- A Substack writer who wants to fire off a Substack Note from inside their writing workflow
- A social media manager checking an account's queue, free slots, and quota for the week
- A user who pastes a Typefully draft link into chat and wants the agent to read and revise that draft
- Users who need Instagram, TikTok, Facebook, or other unsupported platforms—the skill covers only X, LinkedIn, Threads, Bluesky, Mastodon, and Substack Notes
- People without a Typefully account or API key—on auth failure the skill explicitly stops rather than scraping the web UI
- Fully unattended auto-posting setups—publishing is irreversible and public, and the skill requires confirmation first
How do you install this skill?
- Core function depends entirely on the overseas Typefully API (api.typefully.com); mainland-China network reachability is unverified — test connectivity before use.
- Publishing is irreversible and public; ensure the agent obtains explicit user confirmation before drafts:publish / --schedule now.
- SKILL.md claims Node.js 18+ but package. requires >=20; install Node 20+ to avoid incompatibility.
- API keys are stored in plaintext (environment variable or .typefully/config. / ~/.config/typefully/config.); avoid committing them to version control.
- --force-overwrite-comments resolves and strips uncovered comment threads server-side and is unrecoverable via the API; use only after per-item user confirmation.
- This is a static source review without execution; confidence is low.
- Shell / CLI
- Network access
- Local filesystem
Node.js 18+Typefully API key
Generic CLI install (works with Claude Code, Cursor, Windsurf, and other agents):
npx skills add typefully/agent-skillsClaude Code via the plugin marketplace:
/plugin marketplace add typefully/agent-skills
/plugin install typefully@typefully-skillsCursor as a remote rule: Settings → Rules & Command → Project Rules → Add Rule → Remote Rule (GitHub), enter https://github.com/typefully/agent-skills.git
Manual: clone the repo and copy skills/typefully/ into your project's .cursor/skills/ or .claude/skills/ directory.
Then configure the API key:
./scripts/typefully.js setupor set an environment variable:
export TYPEFULLY_API_KEY=your_key_hereHow do you use this skill?
Once installed, send your agent any of these to trigger it:
- Draft a tweet about our new release and schedule it for 9am tomorrow
- Post this to X and LinkedIn:
- Create a thread about AI trends
- Show my scheduled posts and remaining publishing quota
The skill triggers on natural-language requests to draft, schedule, post, or check tweets/posts/threads, or when a Typefully draft URL is pasted. Everything runs through the script and outputs JSON. Typical flow: run setup once to store your API key and default social set (an identity's connected platforms); drafts default to that set, and single-argument commands require --use-default. Publishing is irreversible and public, so the agent must confirm unless you explicitly say "publish now". Planned drafts (--plan) sit on the calendar but never auto-publish; confirm them with drafts:schedule. Examples:
./scripts/typefully.js drafts:create --text "Launch day!" --platform x,linkedin --schedule next-free-slot
./scripts/typefully.js queue:get --start-date 2026-02-01 --end-date 2026-02-29
./scripts/typefully.js drafts:publish 456 --use-defaultWhat are this skill's strengths and limitations?
- Officially maintained, built on the Typefully API, with complete draft/schedule/queue workflows for six platforms
- Zero-dependency Node.js script with JSON output, easy for agents to parse reliably
- Sensible safety design: confirmation before publishing, stop on auth failure, no web-scraping fallback
- Task-specific reference docs for X analytics, LinkedIn mentions, X Articles, and comment workflows
- Hard-bound to a Typefully account and API key; unusable standalone
- Only six platforms supported—Instagram, TikTok, and other major networks are out of scope
- The allowed-tools frontmatter is a Claude Code-style permission extension needing light edits elsewhere
- Platform quirks pile up: Substack Notes rejects threads and conflicts with media/link cards, easy to trip over
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Typefully Social Media Skill this page | 64 · Recommended | ★ 61 | 24d ago | MIT |
| LinkedIn Search Skill | 55 · Use with care | ★ 47k | 17d ago | MIT |
| Job Scraper — Local-First Job Search Skill | 54 · Use with care | ★ 47k | 17d ago | MIT |
| X Article Publisher | 56 · Use with care | ★ 874 | 9mo ago | MIT |
| Crowdcast Multi-Agent Social Simulation | 50 · Use with care | ★ 11 | 5mo ago | — |
The README points to the official Typefully MCP Server (https://support.typefully.com/en/articles/13128440-typefully-mcp-server) as an alternative offering native tool access in Claude Code; choose it for deeper integration, while this skill suits any Agent Skills-compatible agent.
How did FollowSkills review this skill?
Least privilege: allowed-tools is only Bash(./scripts/typefully.js); irreversible public publishing explicitly requires user confirmation; auth failures forbid credential hunting or web-UI fallback; destructive flags like --force-overwrite-comments require explicit confirmation in a separate turn; API key storage and data flows are disclosed; MIT license with clear attribution. Deductions: publish confirmation relies on instruction-following rather than technical enforcement; the script is truncated in evidence so all external calls cannot be fully audited.
Zero-dependency CLI, JSON output, errors with hints/actions, --use-default ambiguity guard, and clear quota/auth failure feedback; repository includes a vitest suite and CI covering key paths. Deductions: static review without execution; script truncated in evidence; environment inconsistency (SKILL.md says Node 18+, package. requires >=20).
Trigger conditions are explicit (drafting/scheduling/publishing social content, dropped Typefully draft URLs), with clear platform boundaries, non-fit cases (x_article standalone, substack single-post) and character limits, plus a phrase-to-action table. Deductions: core function depends entirely on the overseas Typefully API with no assessed mainland-China reachability or fallback; no Chinese-language support declared.
Well-layered docs (SKILL.md plus on-demand references), Keep-a-Changelog format CHANGELOG, freshness check, update paths, license, troubleshooting, and clear maintenance ownership (Typefully's official repo). Deductions: Node version inconsistency (18+ vs >=20), README version badge 1.0.0 without a matching semantic versioning mechanism, and incomplete script visibility.
Goal is clear (agent-side drafting/scheduling/management of social posts), output is structured JSON with abundant directly usable command examples and a phrase-action mapping; marginal value over manual work is evident. Deductions: static review cannot verify real output quality; effectiveness ultimately depends on the Typefully API and paid quota; no real execution output samples.
Repository contains a CI workflow (Node 20/22 matrix running vitest) and committed tests covering key paths such as analytics and comments; claims are locally reproducible via npm test, and changelog entries can be cross-checked against code. Deductions: single-source (official repo self-tests), no independent third-party verification, test files truncated in evidence, no execution performed, limiting static confidence.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →