ADHX — X/Twitter Post Reader
Turns any X/Twitter link into clean, LLM-friendly JSON — including full Article content, author info, and engagement metrics — with no login or browser scraping.
The skill only calls a public adhx.com API via curl — no auth, no credentials, no local writes; the data flow (only username/statusId from a tweet URL) is clearly disclosed, with no red-line risk. Deducted: content is relayed through a third-party server whose retention policy is not addressed at the skill level; no rollback or failure-recovery mechanism.
The workflow (parse URL → curl → use JSON) is self-consistent, with a response schema and basic error guidance ('inform the user if the API errors or returns empty'). Deducted: repo tests (config/processor) do not cover the skill's API path; schema authenticity, missing fields, timeouts and rate limits are unverified and failure feedback is thin.
Trigger conditions are clear (user shares an x.com/twitter.com/adhx.com link to read/summarize), with a URL-format table and examples. Deducted: no declared availability or SLA for the third-party adhx.com API; X content itself is not directly reachable from mainland China and adhx.com reachability is unverified; non-fit cases (protected/deleted posts) are undeclared.
SKILL.md is well structured with metadata source link, schema, example and notes; the repo has an MIT license, release-please versioning and maintenance signals. Deducted: the skill itself carries no version or changelog, the response schema is a unilateral author claim, and API usage limits are undocumented.
The core task (turning a tweet link into structured JSON for LLM analysis) is direct and the command is executable, with clear marginal value over manual scraping. Deducted: static review cannot confirm the live API matches the documented schema; dependency on a third-party service risks shutdown or rate limiting, and output-quality evidence is limited.
Auditable primary material exists (SKILL.md, README, CI, tests) with reasonable fact/inference separation. Deducted: the repo's CI and tests do not exercise the skill's API key path, there is no third-party execution evidence, and schema correctness rests entirely on author claims — not independently reproducible.
- The skill depends entirely on the third-party adhx.com API, which may change, rate-limit, or shut down; it is also the sole core dependency and its reachability from mainland China is unverified — test before relying on it.
- Tweet content is relayed through adhx.com servers; avoid using it with sensitive links, and note the documented response schema is unverified and actual fields may differ.
- This was a static review with no execution; scores are based on source files only and do not constitute empirical evidence of API behavior.
What does this skill do, and when should you use it?
ADHX is an Agent Skill from the itsmemeworks/adhx repository that converts x.com, twitter.com, or adhx.com post links into structured JSON. It calls a public ADHX API endpoint and returns the tweet text, author profile, timestamp, engagement metrics, and the full Markdown content of long-form X Articles. The whole workflow is a single curl command with no authentication and no local dependencies. It suits any agent workflow that needs to read, summarize, or analyze X content.
Parses a shared X/Twitter URL to extract the username and statusId (supporting x.com/{user}/status/{id}, twitter.com/{user}/status/{id}, and adhx.com/{user}/status/{id} formats), then fetches https://adhx.com/api/share/tweet/{username}/{statusId} with curl. The JSON response contains text, author (name, username, avatar), createdAt, engagement (replies, retweets, likes, views), and — for long-form posts — an article object with title, preview text, cover image, and full Markdown content including inline image references, which the model uses for summarizing, analysis, or key-point extraction.
- A user pastes an X link in chat and wants the agent to read and summarize the tweet directly, without opening a browser.
- Analyzing the full body of an X Article: the API's article.content field provides complete Markdown with image references.
- Research or monitoring workflows that need a tweet's engagement numbers (likes, retweets, views) as quantitative input.
- Quickly extracting key points and author context from a tweet for a report or group chat reply.
- A lightweight alternative for agents that cannot run headless browsers but still need to read X content.
What are this skill's strengths and limitations?
- No authentication required — a single curl call with near-zero setup cost.
- Returns clean structured JSON including the full Markdown of long-form X Articles, far better suited to LLMs than page scraping.
- Handles x.com, twitter.com, and adhx.com link formats.
- The skill is just a SKILL.md instruction file with no local scripts, so it is highly portable.
- Fully dependent on the third-party adhx.com API; its rate limits, uptime, and terms are not documented in the source.
- Covers only X/Twitter content — the broader ADHX app also supports Instagram, TikTok, and YouTube Shorts, but the skill exposes no such capability.
- No SLA on quota or stability; on failure the only recourse is telling the user the post is unavailable.
- Accuracy and refresh behavior of engagement metrics (e.g., view counts) are unverified in the documentation.
How do you install this skill?
Claude Code: run /plugin marketplace add itsmemeworks/adhx, then /plugin install adhx. Other skills-compatible agents (Cursor, Gemini CLI, Codex, Copilot):
mkdir -p "$HOME/.cursor/skills/adhx" && curl -sL https://raw.githubusercontent.com/itsmemeworks/adhx/main/skills/adhx/SKILL.md -o "$HOME/.cursor/skills/adhx/SKILL.md"
(Adjust SKILLS_DIR to wherever your agent looks for skills.)
How do you use this skill?
When a user shares an X/Twitter link, extract {username} and {statusId} from the URL path and run:
curl -s "https://adhx.com/api/share/tweet/{username}/{statusId}"
Then answer the user's question from the returned JSON. Example prompt: "Summarize this post https://x.com/dgt10011/status/2020167690560647464". If the API errors or returns empty, tell the user the post may not be available.