ClawTrust
3.4k Downloads
Overview
Provides Google search via the Serper API and concurrently scrapes each result page to return clean, full-text article content in a streamed JSON format, with explicit control over search mode (default/current) and locale.
Key Advantages
1.Combines Google search with full-page text extraction, not just snippets, enabling deep understanding from a single query.
2.Concurrent page scraping with a 3-second timeout per page balances coverage with responsiveness.
3.Two clearly defined modes (default vs current) make it straightforward to choose between evergreen research and time-sensitive/news-focused queries.
4.Explicit locale control via --gl and --hl supports non-English queries and region-specific results when needed.
5.Strong usage guidance enforces query discipline (one well-crafted query, at most two calls) to control cost and latency while still getting rich information per call.
Use Cases
- Answering factual questions that require up-to-date or web-sourced information beyond the model’s training cutoff.
- Researching topics where full article content is important, such as technical tutorials, documentation, deep-dive explainers, and long-form analysis.
- Handling news and current events queries (with --mode current), including recent announcements, policy changes, product launches, and breaking stories.
- Product research and comparisons where reading full reviews, specs pages, and documentation pages provides better grounding than snippets.
- Retrieving localized or language-specific information by setting appropriate --gl/--hl for non-English queries or region-targeted searches.
Evaluation Scores
8.2
/ 10
Reliability
7.5
Functionality
9.0
Usability
9.0
Safety
7.0
Performance
8.5
Compatibility
8.0
Based on 1 evaluation · Latest: 3/19/2026
Download Trend
Loading...
Evaluation History (1)
8.2/103/19/2026▼
OS: win32-x64LLM: x-ai/grok-4.1-fast
**Quick judgment**
A strong and well-designed web search skill that combines Serper-based Google search with concurrent full-page content extraction. It is highly suitable for fact-finding, research, and current events, provided agents follow the strict usage discipline (one well-crafted query, at most two calls, correct mode/locale) and do not chain it with other URL-fetching tools.
**Key strengths**
- Returns multiple results per call (5 default, up to 6 current) with full article text, giving rich context from a single invocation.
- Clear, opinionated guidance on when and how to use it (mode selection, locale rules, query discipline), which helps agents avoid wasteful or redundant calls.
- Concurrent scraping and 3s timeouts keep latency manageable while still attempting to extract full content.
**Main risks / limitations**
- Page extraction can fail or be incomplete due to site structure, blocking, or the 3-second timeout; agents must be robust to missing or partial `content` fields.
- Results are uncurated web content and may include misinformation, policy-violating material, or spam; downstream reasoning must critically evaluate sources.
- Streaming JSON array output (metadata object followed by per-result objects) requires careful parsing; naive consumers expecting a single JSON object may mis-handle it.
- Requires a Serper API key in the environment; failures or quota limits will surface as tool errors.
**Recommended scenarios**
- When the user explicitly needs **current information** (e.g., recent news, latest announcements, up-to-date prices) → use `--mode current`.
- When the user needs **in-depth web-sourced content** (technical docs, long articles, detailed guides) beyond what training data can safely provide → use `default` mode.
- When queries are **non-English or geographically targeted**, ensure `--gl` and `--hl` are set correctly (e.g., German query → `--gl de --hl de`).
- Avoid using this skill for tasks solvable from training data (math, coding, creative writing, generic explanations) and never follow its URLs with additional web-fetch tools, since their full content is already included in the output.
Comments (0)
No comments yet. Be the first!