A precision-engineered web search pipeline for AI agents that minimizes token usage through semantic caching and source validation.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install vryfik-skill
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install vryfik-skill using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Searching Precisely?
Searching Precisely is a sophisticated middleware for AI coding agents designed to drastically reduce the cost and latency of web-based information retrieval. By implementing a local intent classification layer and a robust semantic cache, it ensures that the agent only performs expensive web searches when truly necessary. This skill is essential for developers looking to build high-performance Openclaw Skills that require up-to-date data without exhausting token budgets.
The skill shifts the heavy lifting of pre-processing and post-processing to local Node.js scripts, leaving the host agent to handle only the raw network requests. This architecture provides a layered security model, including SSRF protection and atomic cache writes, ensuring that data retrieval is both safe and efficient.
Searching Precisely Use Cases
- Reducing API costs by serving repetitive documentation queries from a local semantic cache.
- Verifying the credibility of technical sources and API specifications before presenting them to the user.
- Performing targeted searches for current events, pricing, or news with strict token budget constraints.
- Orchestrating multi-query searches to resolve ambiguous user intent into precise technical answers.
How Searching Precisely Works
- Intent Classification: The system parses the user query locally to determine if a web search is actually required.
- Budget Initialization: A budget controller sets strict token limits for the upcoming operation to prevent runaway costs.
- Semantic Cache Check: It searches a local database for high-similarity previous results; if a match is found, the pipeline terminates early with the answer.
- Query Rewriting: If the cache misses, the query is optimized into targeted sub-queries for better search engine performance.
- Host-Delegated Search: The host agent performs the actual web search based on the refined queries.
- Credibility Probing: Source URLs are validated via HEAD requests to check availability and reputation without downloading full page bodies.
- Assembly & Caching: Validated fragments are assembled into a coherent answer, which is then written back to the semantic cache for future use.
Searching Precisely Setup
To integrate Searching Precisely into your environment, ensure you have the Node.js runtime installed and follow these steps:
# Ensure the cache directory exists with restricted permissions
mkdir -p ~/.antigravity/search-cache/
chmod 700 ~/.antigravity/search-cache/
# The skill utilizes local scripts via the shell tool
# No additional npm installation is required if using the Openclaw bundle
Searching Precisely Data Schema & Taxonomy
The skill manages its local state and intelligence using a structured file system and JSON-based messaging:
| Component | Path / Format | Description |
|---|---|---|
| Semantic Cache | ~/.antigravity/search-cache/index.json |
Stores query vectors and successful answer fragments (mode 0o600). |
| Intent Output | JSON {intent, confidence} |
Communication between the parser script and the host agent. |
| Probing Results | JSON {verdict, available} |
Results of the parallel credibility probe for source URLs. |
| Budget State | JSON {state.remaining} |
Tracks the remaining token budget across the pipeline lifecycle. |
Searching Precisely Advanced Features
- Parallel Credibility Probing: Simultaneously validates multiple source URLs using HEAD-only requests to minimize latency and data transfer.
- Two-Layer SSRF Protection: Combines hostname pattern checking with DNS-resolved IP verification to prevent rebinding attacks.
- Atomic Cache Management: Uses temp-file-and-rename logic to ensure the local search cache never becomes corrupted during write operations.
- Coherence Scoring: Automatically flags assembled answers if the underlying data fragments lack logical consistency or completeness.
SKILL.md
Loading
Related Openclaw Skills
Generate Facebook Ad Creative Variations
An automated AI skill for generating diverse, audience-specific Facebook ad copy and visual assets for optimized campaign performance.
Maxxit 0G Trading
A decentralized AI trading skill that enables portfolio-aware trade decisions via 0G compute and ZK-verified alpha sharing across multiple DEXs and global equity markets.
Generate Finance Ad Creative Brief
An AI-driven workflow designed for finance teams to plan, draft, and refine high-converting ad creative briefs and visual hooks.
Iteration Optimizer
A specialized skill for Openclaw Skills that automatically refines AI-generated code through iterative evaluation and scoring.
Generate Estate Planning Law Firm Client Education Handout
An AI-powered tool for estate planning attorneys to generate professional client education materials, including visuals and FAQ guides.
Generate Developer Ad Creative Brief
Automate the creation of strategic ad creative briefs specifically tailored for developers, technical founders, and product engineers.
Featured*
