Internet Lookup Verifier

Verify information by performing an internet lookup before answering questions.

Security Scan

The skill's instructions, requirements, and behavior are consistent with its stated purpose (performing web lookups to verify answers); it requests no credentials, has no install steps, and stays within expected scope.

Purpose & Capability

The name/description (verify answers via internet lookup) aligns with the instructions: use web_search and web_fetch, extract title/URL/snippet, evaluate sources, and return a short answer with references. No unrelated binaries, env vars, or installs are requested.

Instruction Scope

SKILL.md confines actions to search/fetch/evaluate web results. It does not instruct reading local files, environment variables, or sending data to external endpoints beyond retrieving public web pages. The trigger language is reasonably scoped (keywords like 'verify', 'source'), though triggers depend on the agent/system prompt.

Install Mechanism

No install spec and no code files — instruction-only skill. This minimizes risk from arbitrary code or disk writes.

Credentials

No required environment variables, credentials, or config paths are declared or referenced in SKILL.md. The skill's operations (search/fetch) do not need additional secrets, so requested privileges are proportional.

Persistence & Privilege

always is false and the skill is user-invocable; it does not request permanent installation or elevated system privileges. It relies on platform-provided web_search/web_fetch tools, which is normal for this capability.

Assessment

This skill is instruction-only and coherent with its purpose: it performs searches and returns cited sources. Before enabling, confirm that your platform's web_search/web_fetch implementations are trustworthy and respect privacy (to avoid fetching sensitive, link-tracking, or malicious pages). If you prefer tighter control, ensure the agent's invocation policy limits when the skill can run or require user confirmation before it performs external fetches.

Purpose

When a user asks a factual question, this skill performs an online search to confirm the answer and returns a concise verified response. It ensures that the assistant’s reply is backed by current, reliable sources.

How it works

1. Trigger – The skill activates when the system prompt or another skill signals a need for verification. Typically this happens before answering a question that contains keywords like fact, true/false, verify, or source.
2. Search – Uses the web_search tool to retrieve up to 5 relevant results (default 3). The query is the user’s question or a re‑phrased version suitable for search engines.
3. Extract – From each result, the skill pulls the title, URL and snippet. If needed, it can fetch the full page with web_fetch for deeper analysis.
4. Evaluate – The assistant checks that at least one source explicitly supports the answer. If sources disagree or none are found, it reports uncertainty.
5. Respond – Returns a short verified answer followed by a list of URLs (and optionally snippets) that back the claim.

Usage example

User: Is the capital of Australia Canberra?

Assistant (using this skill): Yes, Canberra is the capital of Australia. Source: https://www.britannica.com/place/Canberra-Australia

Limitations

• Relies on the quality of public search results.
• May not handle niche or very new information if it hasn’t indexed yet.
• Does not guarantee 100 % accuracy; it only indicates that sources support the claim.

Extensibility

Add a references/ folder with detailed guidelines for interpreting ambiguous results or handling conflicting sources. The skill can be extended to use more advanced NLP techniques for source credibility scoring.