forge-web-search

What problem does it solve?

This Skill prevents AI coding agents from treating unverified web snippets as facts by enforcing disciplined search, source triage, bounded fetching, and citation hygiene.

Core Features & Use Cases

• Query construction and iteration: Prefer specific terms, exact phrases, site/date filters, and iterative refinement based on results.
• Credibility-based source triage: Rank sources with an explicit credibility hierarchy and skip SEO-spam and AI-laundered content.
• Fetch-vs-snippet decisions with budgets: Avoid fetching unless needed, enforce caps on search and fetch loops, deduplicate visited URLs, and respect robots/rate limits.
• Citation and anti-hallucination safeguards: Require verifiable URL resolution, exact quote matching, and access/publication date notes.
• Constrained synthesis with conflicts surfaced: Synthesize from noise-filtered sources and highlight contradictions when they appear.
• Worked agent loop and verifier: Includes an end-to-end reference loop and a shell verifier that flags unbounded loops and unchecked quoting.

Quick Start

Ask your agent to use forge-web-search when it needs online research, such as verifying Postgres documentation details before writing code or explaining an API behavior.