Internet Skill

Deep iterative web research for current facts, verification, due diligence, comparisons, market maps, and cited synthesis.

internet turns broad research prompts into a disciplined investigation loop: preflight the question, plan source classes, search in passes, read primary evidence, check contradictions, track claims, and synthesize with citations. When research should become durable memory, it prepares review-gated AKBP source and claim previews.

Install

Install this repository with any skill-capable agent or skill manager that can consume a skills/ directory. The skill itself lives at skills/internet/SKILL.md and is intentionally agent-portable.

With SkillKit:

npx skillkit install rohitg00/internet-skill

With an agent-specific plugin marketplace, add rohitg00/internet-skill and install the internet skill using that agent's normal plugin flow.

What It Does

• Plans research before searching.
• Catches ambiguous prompts, generic rankings, numeric traps, temporal traps, and high-stakes edge cases.
• Searches orientation, primary, independent, community, contradiction, and gap-fill passes.
• Tracks claims, source tiers, dates, confidence, caveats, contradictions, and unresolved gaps.
• Prepares AKBP source and claim previews without silently writing durable memory.
• Produces cited briefs, reports, comparisons, timelines, recommendation memos, and evidence packs.
• Includes a workspace helper for long-running investigations.

Example Prompts

Use $internet to research whether we should adopt Bun for a TypeScript monorepo in 2026.

Use $internet to build a detailed competitor map for AI browser automation tools.

Use $internet to verify the latest Kubernetes Gateway API production readiness claims.

Workspace Helper

WORKSPACE="$(python3 skills/internet/scripts/research_workspace.py create "AI browser automation tools" --depth deep | sed -n '1p')"
python3 skills/internet/scripts/research_workspace.py add-query "$WORKSPACE" --pass-label orientation --query "AI browser automation tools 2026" --tool web --why "discover landscape" --result "identified primary projects"
python3 skills/internet/scripts/research_workspace.py add-source "$WORKSPACE" --url "https://example.com" --title "Source title" --type primary --tier 1 --published YYYY-MM-DD
python3 skills/internet/scripts/research_workspace.py add-claim "$WORKSPACE" --claim "Material claim" --evidence "Evidence note" --source-id S001 --confidence High
python3 skills/internet/scripts/research_workspace.py add-evidence "$WORKSPACE" --source-id S001 --claim-id C001 --evidence "Structured evidence note"
python3 skills/internet/scripts/research_workspace.py summary "$WORKSPACE"

The helper creates a working folder and records query, source, claim, evidence, contradiction, gap, and AKBP intake ledgers. Recorder commands also update research.md, so deep research does not end with an empty-looking report.