JPeetz/agent-skills / skills/documentation-content/read-x-articles
read-x-articles
VerifiedUse this skill when a user drops an x.com or twitter.com URL (especially an X Article like /i/article/ID) and expects it read, or says "read this X post/article/thread" in any AI coding agent. Resolves the canonical /i/article/ID URL (which serves plain text) instead of giving up on the login/JS wall that hits profile and /status/ pages. Includes browser-tool and API fallbacks.
| File | Kind | Size |
|---|---|---|
| CHANGELOG.md | markdown | 742 B |
| LICENSE | other | 1.0 KB |
| SKILL.md | skill md | 5.2 KB |
| evals/evals.json | config | 1.9 KB |
| references/url-forms.md | markdown | 1.9 KB |
| scripts/read_x_article.py | script | 2.5 KB |
SKILL.md
View on GitHub---
name: read-x-articles
description: >-
Use this skill when a user drops an x.com or twitter.com URL (especially an
X Article like /i/article/ID) and expects it read, or says "read this X
post/article/thread" in any AI coding agent. Resolves the canonical
/i/article/ID URL (which serves plain text) instead of giving up on the
login/JS wall that hits profile and /status/ pages. Includes browser-tool
and API fallbacks.
version: 1.0.0
author: Hermes Agent
license: MIT
compatibility: >
Cross-platform: Claude Code, OpenAI Codex, GitHub Copilot, Cursor, Windsurf,
Gemini CLI, OpenClaw, Hermes Agent, and any SKILL.md-compatible agent.
tags:
- x
- twitter
- articles
- web-extract
- reading
- long-form
platforms:
- claude-code
- codex
- cursor
- gemini-cli
- openclaw
- hermes-agent
---
# Read X (Twitter) Articles
Turn any X link a user shares into the full article text. Do NOT default to "I
can't read X" — X long-form Articles DO extract cleanly via the right URL.
## When to Use
- A user drops an `x.com/...` or `twitter.com/...` link and expects you to read it.
- The link is an X **Article** (long-form essay/interview) — canonical path
`/i/article/<ID>`.
- You need primary source text from an X essay/thread as material for work
(e.g. writing, video scripts, analysis).
- Anyone claims "X content can't be read" — that is the signal to trigger, not
capitulate.
## The core insight (learned 2026-08-12)
- X **Articles** are readable end-to-end at the canonical
**`https://x.com/i/article/<ID>`** **when fetched by a JS-executing
web-extractor / browser tool** — that returns the FULL body, not a snippet.
- **Bare HTTP fetch (`urllib`/`curl`) of the article often returns a login/JS
shell.** The canonical URL is necessary but a plain request is NOT sufficient
— you need a render-capable tool. This is why `web_extract` (SE/agent-grade)
succeeds where `curl` fails.
- **`/status/<id>` posts and profile pages** are JS-rendered and hit the
login wall → they come back empty even to a browser. That is the apparent
"can't read X." The fix is resolving the article URL AND using the right
fetch tool, not giving up.
- **Images matter.** The article's cover, diagrams, charts, and screenshots
can be essential to understanding it. Extract the media URLs and analyze
them (vision) as part of the read — do not stop at the text.
## Steps
1. **Try first, judge later.** When given any X link, immediately attempt it
with a web-extract/browser tool. Declaring it unreadable *before* trying is
the exact mistake this skill prevents.
2. **Use the right fetcher:** prefer the agent's `web_extract`-style tool or a
browser render — NOT bare `curl`/`urllib`. If the only tool is a plain HTTP
client, expect a login shell and fall back to a browser.
3. **If it's a `/status/` URL or returns a login/JS shell**, resolve the
canonical article URL:
- The article ID often appears in the status link's query/path when the
post is an article share.
- Find the author's `/i/article/<ID>` share link on their profile, or ask
the user for the article link directly.
- Then fetch `https://x.com/i/article/<ID>` with a render-capable tool.
4. **Validate the body**: a real extract contains substantive paragraphs, NOT
just "Log in or sign up for X." A 1,900+ word essay returns complete.
5. **Fallbacks if a render-capable extract still fails**:
- Use a **browser tool** (CDP-style) to render the JS page, then read the
rendered text/DOM.
- If the account has X API access (e.g. an `xurl`-style CLI), query the
article tweet and read `data.article.plain_text`. A dead/unauthed CLI
token is NOT a blocker — prefer web/browser first.
6. **Use the content.** It's legitimate primary source. Quote it and cite the
author. If it's a community take (e.g. a power-user writing about a product),
it's still real material for summaries, scripts, and analysis.
7. **Read the images too.** After extracting the body, the text extractor may
drop inline images (screenshots, diagrams, UI captures) that are essential to
understanding the article. **Fetch the raw HTML of the page** (curl or similar)
and extract all `pbs.twimg.com/media/` URLs — these are the article's inline
images. Analyze the meaningful ones (not user avatars) with a vision tool.
Report any visual info that adds to the text: screenshots of a UI, diagrams
that explain a workflow, charts with data, or a cover that sets the tone.
Do not skip this — images are part of the article.
## Pitfalls
- Do not assume "articles are paywalled/login-walled." The canonical
`/i/article/<ID>` endpoint serves plain text; the wall applies to the
JS-rendered profile/status pages, not this endpoint.
- A dead X API token (401) is unrelated to article reading — do not block the
task on fixing xurl/auth when web extraction works.
- Verify you got the actual article (thesis + body), not just a title + login
prompt, before treating it as read.
## Verification
- Did the extraction return the article's real prose (title, paragraphs,
thesis) rather than a shell? If yes, it's read.
- Report the exact URL you read + a short grounded summary so the user can
verify.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Unverified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |