← Back to Skills Marketplace
62
Downloads
0
Stars
0
Active Installs
8
Versions
Install in OpenClaw
/install vryfik-skill
Description
Use when searching the web, documentation, or current information where token efficiency matters. Triggers on queries about API docs, current events, pricing...
Usage Guidance
This skill appears internally coherent and implements a local pipeline (intent parsing, rewriting, caching, credibility probes, assembly) that delegates actual page retrieval to the host agent. Before installing:
- Ensure you trust the host agent's 'web_search' tool (it performs the live GETs and supplies fragments to the skill). The skill assumes URLs come from a trusted caller.
- Accept that queries (cache keys/snippets) will be stored under ~/.antigravity/search-cache/ (SKILL.md states files created with 0o600 and dir 0o700). If that concerns you, review or relocate the cache directory before use.
- Note the skill issues outbound HEAD requests (no bodies) for availability checks; network egress must be acceptable.
- No secrets/credentials are requested by this skill.
If you need higher assurance, review the bundled scripts (they are small, readable JS) and verify the host's web_search implementation sanitizes user-supplied URLs and does not forward sensitive local/internal URLs to the skill.
Capability Tags
Capability Assessment
Purpose & Capability
Name/description (token‑efficient web/document search) matches the artifacts: intent parsing, query rewriting, budget control, cache, credibility probes, and assembly. Required runtime (node) is appropriate and no unrelated credentials or binaries are requested.
Instruction Scope
Instructions limit network activity to host-provided web_search for content retrieval and the skill's parallel-probe script (HTTP HEAD) for availability checks. The skill reads/writes a local cache (~/.antigravity/search-cache/) and uses local data files (intent patterns, domain reputation). Cache may contain user query text (SKILL.md documents this). Review host agent behavior: host provides full GET/search results to the skill for post-processing, so the security posture depends on trusting the host tool to sanitize user-supplied URLs.
Install Mechanism
No install script; code files are included and executed via the shell tool using node. No remote downloads, package installs, or URL fetch/install steps are present.
Credentials
The skill requests no environment variables or credentials. Local file access is limited to a single cache directory and bundled data files; network access is restricted to outbound HEAD probes (parallel-probe) and relies on the host for GET/search calls—this is proportionate to the stated functionality.
Persistence & Privilege
always is false and disable-model-invocation is set to true (skill cannot invoke the model autonomously). The only persistent artifact is the local cache (~/.antigravity/search-cache/) which the SKILL.md documents and the scripts write with restricted file modes (0o600/0o700). No configuration or credential changes to other skills are made.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install vryfik-skill - After installation, invoke the skill by name or use
/vryfik-skill - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.7
v1.0.7 is a metadata refinement release.
- Added explicit "metadata.openclaw.requires.bins" field to declare required binaries.
- Moved the "required binaries" comment into the formal YAML manifest.
- No changes to skill logic, pipeline, permissions, or instructions.
- No code or functionality alterations; update is manifest/metadata only.
v1.0.6
v1.0.6 — Minor manifest YAML corrections
- Fixed typo in the manifest key from disable_model_invocation to disable-model-invocation
- Changed "required_binaries" key to "required binaries" for consistency
- No other changes to instructions, logic, or security notes
v1.0.5
v1.0.5 is a policy/security manifest update:
- Added disable_model_invocation: true to enforce non-autonomous use via OpenClaw policy.
- Introduced required_binaries with "node" to specify required runtime.
- No changes to files or functional pipeline; documentation/manifest only.
v1.0.4
v1.0.4 introduces enhanced environment and security detail clarifications.
- Declared Node.js (>=18) as an explicit dependency for host agent compatibility.
- Expanded IO section with details that cache directory is auto-created.
- Upgraded SSRF protection to a two-layer system: (1) hostname pattern check and (2) DNS lookup with post-resolution IP validation.
- Updated security notes to reflect layered SSRF checks and reinforce DNS rebinding prevention.
- No changes to core pipeline or behavior.
v1.0.3
v1.0.3 introduces enhanced security controls and clarifies agent/skill responsibilities.
- Now explicitly requires script invocation via shell tool; no direct user-script execution.
- Adds SSRF protection: private IPs, localhost, and link-local addresses are blocked before probing.
- Clarifies that web search requests (GET/API calls) are delegated to the host agent, not performed by this skill.
- Documents required invocation order, early exits, and precise JSON argument passing for each step.
- Updates security manifest with new `autonomous: false`, `host_provides`, and `trust_model` fields.
- Improved security notes and operating instructions for safer, more auditable use.
v1.0.2
- Major update: skill focus shifted from codebase search to web and documentation search only.
- Removed support for local file system operations and shell/subprocess execution; now web-only and subprocess-free.
- Dropped dependency on external binaries such as ripgrep.
- Updated security manifest to reflect reduced permissions and no shell access.
- Token budgets and pipeline simplified; no layered local search, now purely cache-then-web-search with credibility checks.
- Streamlined documentation and usage examples to match the new web-focused architecture.
v1.0.1
**Added explicit security policy and technical manifest to SKILL.md**
- Added ClawHub Security Manifest section in YAML frontmatter, declaring all permissions, dependencies, I/O scopes, and security notes.
- Frontmatter now specifies which scripts have local read/write, network, and shell permissions, and their boundaries.
- Added details about ripgrep dependency for layered-fetcher.js.
- Included security best practices: atomic cache writes, safe path handling, restricted network access, and no code generation.
- Updated ClawHub Security Notes to direct users and static analyzers to use the manifest for authoritative metadata.
- No changes to pipeline logic or user-facing search behavior.
v1.0.0
Initial release of the searching-precisely skill for token-efficient search in codebases and documentation.
- Introduces a hierarchical search pipeline with early exits to minimize token usage.
- Implements semantic caching, layered fetching (local grep, fragment, web), and parallel credibility checks.
- Provides clear rules for layer selection and query handling.
- Includes utility scripts for intent classification, query rewriting, budget management, validation, and cache management.
- Emphasizes best practices to avoid unnecessary token or I/O costs.
- Details security measures and scope-limited I/O for each component.
Metadata
Frequently Asked Questions
What is vryfik skill?
Use when searching the web, documentation, or current information where token efficiency matters. Triggers on queries about API docs, current events, pricing... It is an AI Agent Skill for Claude Code / OpenClaw, with 62 downloads so far.
How do I install vryfik skill?
Run "/install vryfik-skill" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is vryfik skill free?
Yes, vryfik skill is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does vryfik skill support?
vryfik skill is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created vryfik skill?
It is built and maintained by briefness (@briefness); the current version is v1.0.7.
More Skills