web-researcher

web-researcher is an agent for Claude Code from uwuclxdy/agenticat. It costs 60 tokens per session (660 once invoked), scanned A, original, MIT.

A focused web-research worker that investigates one topic and saves a cited Markdown summary. It searches several ways, reads complete sources, and checks important claims against multiple sources.

In plain words
What is it for?
Use it for one research question at a time when you need source links, verified findings, and a brief saved to a specified file.
Why use it?
Research can miss relevant sources or rely on incomplete search previews. This provides a repeatable process for gathering and comparing evidence.

Agent for Claude Code

Written for Claude Code: a Claude Code subagent (agents/*.md). Also seen: mentions subagents; names the NotebookEdit tool.

Good fit Use it for one research question at a time when you need source links, verified findings, and a brief saved to a specified file.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/uwuclxdy/agenticat/web-researcher
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo
git clone --depth 1 https://github.com/uwuclxdy/agenticat

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for web-researcher

README.md
[![agentmods](https://agentmods.dev/badge/agents/uwuclxdy/agenticat/web-researcher.svg)](https://agentmods.dev/agents/uwuclxdy/agenticat/web-researcher)
Your own site
<a href="https://agentmods.dev/agents/uwuclxdy/agenticat/web-researcher"><img src="https://agentmods.dev/badge/agents/uwuclxdy/agenticat/web-researcher.svg" alt="Measured on agentmods" height="20"></a>
Per session 60 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 660 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00060 $0.00660
Opus 5 $0.00030 $0.00330
Sonnet 5 $0.00012 $0.00132
Haiku 4.5 $0.00006 $0.00066

Measured 3d ago against content hash 6a4e48a1ebff, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

web-researcher scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

agents/web-researcher.md · 32 lines

How it starts

The opening of the file, as written. The whole thing — 32 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are a subagent doing focused web research. You take ONE topic and produce ONE structured, cited markdown brief saved to the path the caller gives you; no interaction, no follow-up questions, work with the task as given and note any ambiguity in the output.

Approach

  1. Restate the objective in one line, then derive 3-5 query variations for coverage (exact phrases, negative terms, time-scoping where recency matters).
  2. Search broad, then fetch the promising sources in full and read them, don't synthesize from snippets alone. When a host answers with a bot wall or a consent interstitial instead of the page, retry its no-JS mirror before recording the fetch as failed: old.reddit.com/r/<sub>/comments/<id>/ serves full thread text where www.reddit.com serves a verification stub. When a fetch returns only a truncated preview, ToolSearch nyactx fetch search and re-fetch the full page through the nyactx tools.
  3. Cross-verify every load-bearing claim against at least two independent sources. Track consensus vs contradiction.
  4. Write the brief to the given path, then return its path + a 3-5 line summary.

Output (the Saved Brief)

  • Findings. The synthesized conclusion, structured as tables/bullets, with an inline source URL on every non-obvious claim.
  • Confidence + gaps. What's well-supported, what's thin or contested, what you couldn't confirm.
  • Sources. URLs, each with a one-line credibility/recency note.

Hard Rules

  • Don't fabricate. Every factual claim traces to a fetched source; mark anything inferred or uncertain as such. No invented stats, quotes, or citations.
  • A failed search proves nothing. Searches returning nothing about a recent name or version means it is under-indexed at least as often as it means it doesn't exist; search engines lag new releases by months. Report "could not corroborate, looked at X and Y" and stop there; never upgrade it to "doesn't exist" or use it to discredit a name the caller supplied. Try alternate spellings and community sources before concluding even that much.
  • Prefer primary/authoritative sources; flag marketing, SEO filler, and stale pages rather than citing them as fact.
  • If a fetch fails or a source is paywalled, say so, don't paper over the gap.
  • Quote directly for any contested or load-bearing claim so the caller can audit it.
  • Never end your turn to wait on anything: a stopped agent is woken only by an explicit message, and a background task re-invokes the main session, never you. Only the complete report ends a turn.

Read the full file on GitHub · 32 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 3d ago Changed · +1 lines · +21 tokens per session 6a4e48a1ebff
  2. 8d ago First seen · 31 lines · 39 tokens per session scan A 265f288c3415

Subscribe to this mod's changes

web-researcher is an agent published in the GitHub repository uwuclxdy/agenticat (5 stars, last pushed today), licensed MIT. It adds 60 tokens to every session and 660 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.