Vajra

Analyze URLs, YouTube videos, tweets, or text for quality, bias, and reliability using the Vajra API (vajra.to). Use when the user asks to fact-check, analyze, evaluate, score, or assess the quality/bias/reliability of any web content, article, video, tweet, or text. Also use when asked to "vajra" something or run content through a signal filter.

Vajra - Signal Filter for the Internet

Vajra is a content-analysis API by Humanity Labs. It scores content for epistemic quality (1-10), detects bias, extracts key takeaways, flags questionable claims, and produces structured verdicts.

Privacy and Data Handling

Important: This skill sends content to the Vajra API at https://www.vajra.to. Before using it:

• Content you submit (URLs, text) is transmitted to Vajra's servers for analysis.
• Analyzed content is cached server-side. If another user submits the same URL, they receive the cached result (0 credits). The original text you submit is not stored -- only the analysis output.
• Every analysis generates a public permalink (e.g. vajra.to/a/ID). These are shareable and publicly accessible.
• Do not submit private, proprietary, or personally identifiable content unless you understand and accept this behavior.
• Vajra's privacy policy and terms are available at vajra.to.

Setup

An API key is required. Get one free at vajra.to/dashboard (Connections tab).

The key must be stored as the environment variable VAJRA_API_KEY. The agent will be prompted to set this during installation.

Analyze Content

bash

curl -s -X POST https://www.vajra.to/api/analyze \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $VAJRA_API_KEY" \
  -d '{"content": "URL_OR_TEXT", "type": "url"}'

Set type to "text" for raw text (max 50,000 chars). Default is "url".

Supported content: articles, YouTube videos, X/Twitter posts, and raw text.

Parse the Response

The response JSON contains:

text

success        - boolean
cached         - boolean (0 credits if true)
credits_used   - 0 or 1
url            - permalink to full report (e.g. https://www.vajra.to/a/UUID)
analysis.title - content title
analysis.quality_score - 1-10 rating
analysis.bias_level    - bias assessment
analysis.markdown      - full report in markdown
analysis.metadata.tldr - one-sentence summary
analysis.metadata.verdict      - reliability assessment
analysis.metadata.key_takeaways - array of takeaways
analysis.metadata.warnings      - array of warnings

Present Results

When showing results to the user, format as:

text

**Title** - Quality: X/10
TLDR: [tldr]
Verdict: [verdict]
Key takeaways: [list]
Warnings: [list if any]
Full report: [permalink url]

Retrieve Existing Analysis

Fetch a previously completed analysis by ID (no auth needed, public endpoint):

bash

curl -s https://www.vajra.to/api/analysis/ANALYSIS_ID

Credits and Pricing

• Cached results cost 0 credits (same URL already analyzed by any user)
• Free tier: 5 analyses/month
• Pro: 100 analyses/month ($12/mo or $79/yr)
• Analyses take 10-60 seconds depending on content length