What does this skill do, and when should you use it?
Taisly Agent Kit is a free, AI-first publishing toolkit shipped as an npm package (@taisly/agent) containing a CLI, SDK, Agent Skill, and MCP server. It lets coding agents publish local or remote videos to connected social accounts after explicit user confirmation, with scheduling and cross-platform reposting built in. Installation is free with no credit card required; paid plans apply only at higher scale. The skill enforces safety rules: validate before posting, confirm before publishing, and never expose the API key.
Once installed it provides the taisly CLI and a local stdio MCP server: check auth status, list connected platform accounts, read per-platform field schemas, preflight-validate local videos, create immediate or scheduled posts, check status by historyId, and repost across platforms. A remote MCP endpoint (https://app.taisly.com/mcp, streamable HTTP + OAuth) accepts only public videoUrl values; local files require the local MCP server or CLI. Every command returns JSON so agents can parse results directly.
- A developer using Codex or Claude Code who wants to auto-publish a demo video to multiple platforms after each build
- A content team using Cursor or a custom agent to schedule captioned short videos to TikTok and Reels
- An automation operator cross-posting an already-published video to other accounts via reposts:create
- An agent workflow that must validate video format and platform constraints before submission to avoid rejections
- A solo creator who wants the agent to handle login, account connection, and publishing without hand-writing platform integrations
How do you install this skill?
- All videos and credentials flow through a third-party SaaS (taisly.com); the publisher is an unverified, unknown identity — assess data-exfiltration risk before posting sensitive content.
- Core function depends entirely on the overseas service app.taisly.com; mainland-China reachability is undisclosed and may be unavailable.
- The repo has no test suite and CI only performs syntax checks; this review is a static source read with no execution.
- The skill triggers real posts to live public social accounts — confirm destination accounts, captions and schedule before use; free-tier limits may push users toward paid upgrades.
- Shell / CLI
- Network access
- Local filesystem
- MCP Server
Node.js@taisly/agent (npm)Taisly account credential or TAISLY_API_KEY
npm install -g @taisly/agent (or run without installing via npx @taisly/agent help). The skill lives in SKILL.md at the repo root and works with Claude Code, Codex, Cursor, OpenClaw, and other agents; Codex users can use the bundled .codex-plugin/plugin. manifest. For auth, prefer the browser flow: taisly setup --agent <agent-slug>, complete the login URL, then taisly checkin --agent <agent-slug>; or manually export TAISLY_API_KEY="taisly_...". The repo does not document steps for copying SKILL.md into a specific skills directory.
git clone --depth 1 https://github.com/taisly/agent.git ~/.claude/skills/agentGenerated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Recommended flow: auth:status → platforms:list → platforms:schema → posts:validate → user confirmation → posts:create → posts:status. Example: taisly posts:create --video ./launch.mp4 --platforms id1,id2 --description "Launch day" --scheduled "2026-06-14T09:00:00+07:00". Over MCP, use taisly_platforms_list, taisly_posts_validate, taisly_posts_create (which requires confirmed: true), etc. You can also write a campaign. payload and pass it with --.
What are this skill's strengths and limitations?
- One integration covers TikTok, Instagram, YouTube, X, and Facebook
- All commands return JSON, so agents parse results without scraping terminal text
- Built-in safety rules: explicit pre-publish confirmation, never print the API key, persist historyId for tracking
- Both a local stdio MCP server (reads local files) and a remote OAuth MCP endpoint
- Free to install and start; no credit card required
- posts:status reads recent history; no dedicated single-post status endpoint yet
- posts:validate is local preflight only; final validation happens server-side in Taisly
- Remote MCP cannot read local video files — public videoUrl only
- Blindly retrying posts:create after a timeout risks duplicate posts; check History first
- Hitting plan limits requires a paid upgrade; the source does not state free-tier quotas
- No test suite or published evidence of verified per-platform posting is shown in the source
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 |
|---|---|---|---|---|
| Taisly Social Media Posting Skill this page | 50 · Use with care | ★ 219 | 22d ago | MIT |
| NotebookLM Automation | 53 · Use with care | ★ 20k | 11d ago | MIT |
| Node Modules Inspector — Dependency Audit Skill | 67 · Recommended | ★ 3k | 1mo ago | MIT |
| OfficeCLI Skill | 38 · Not recommended | ★ 32k | 12d ago | Apache-2.0 |
| Engram Persistent Memory Protocol | 68 · Recommended | ★ 7.2k | 11d ago | MIT |
How did FollowSkills review this skill?
SKILL.md mandates explicit user confirmation before publishing, forbids printing TAISLY_API_KEY, discourages blind retries and mass posting — good least-privilege and confirmation posture. Deducted because all data (videos, credentials, publish actions) flows through an unverified third-party SaaS (taisly.com), the local credential sits in ~/.taisly/config. with no isolation/rollback/disclosure of the data flow beyond basics.
Internally self-consistent: commands, MCP tools, error codes, and workflow steps align; an error-handling section gives concrete feedback paths for abnormal input. Deducted for absence of tests — CI only runs node --check syntax checks, so key paths cannot be reproduced from static evidence.
Audience, scenarios, non-fit boundaries ('What this package is not') and trigger conditions are clearly declared. Deducted because core function depends entirely on app.taisly.com, an overseas service whose mainland-China reachability is undisclosed — a risk for FollowSkills' Chinese user base.
Version (0.2.8), MIT license, known-limitation disclosure, multilingual README, coherent npm/server. metadata. Deducted for unverified publisher identity, no CHANGELOG or explicit maintenance/update commitment, and placeholder platform IDs in examples.
The core path (discover → validate → confirm → create → status) is complete, JSON output is agent-consumable, and local preflight validation reduces failure cost. Deducted for no execution evidence that outputs are directly usable, and limited evidence of marginal benefit over alternatives; statically capped below 7.
Multiple metadata sources (MCP Registry, npm, Glama) and auditable repo files exist. Deducted because all effectiveness claims (successful publishing, platform connectivity) lack third-party execution evidence or reproducible tests; CI is syntax-only, statically capped below 5.
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 →