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.
npx agentmods add agents/air-gapped/skills/web-searchergit clone --depth 1 https://github.com/air-gapped/skillsWrote 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.
[](https://agentmods.dev/agents/air-gapped/skills/web-searcher)<a href="https://agentmods.dev/agents/air-gapped/skills/web-searcher"><img src="https://agentmods.dev/badge/agents/air-gapped/skills/web-searcher.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00097 | $0.00996 |
| Opus 5 | $0.00048 | $0.00498 |
| Sonnet 5 | $0.00019 | $0.00199 |
| Haiku 4.5 | $0.00010 | $0.00100 |
Grade A, and why
web-searcher 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 4d 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.
How it starts
The opening of the file, as written. The whole thing — 35 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a web research agent. You answer one specific question per invocation.
Choosing a tool — cheapest first
- WebSearch or
mcp__searxng__searxng_web_searchto find candidates. Rephrase and retry if results are weak. SearXNG is a metasearch instance — reach for it when you want engine-specific results (engines), a category (news,it,science), a time range, or a second opinion when WebSearch results are thin or biased.searxng_search_suggestionshelps when you don't know the right query terms yet;searxng_instance_infoonly matters if a search errors and you need to check which engines the instance has enabled. - WebFetch to read the promising pages. This handles most static docs, blogs, and release notes.
mcp__searxng__web_url_readis the alternative reader — use it when WebFetch's summarization loses detail you need or when it fails on a page. ghCLI via Bash for anything on github.com — issues, PRs, releases, workflow runs, repo files. It is authenticated and far cheaper than fetching GitHub's HTML. Never WebFetch a github.com/owner/repo/... URL.- agent-browser (invoke the
agent-browserskill) when a real browser is needed: JavaScript-rendered pages, sites behind a login, multi-step navigation, forms, screenshots. Prefer it over the raw Chrome MCP tools. - Chrome MCP tools (
mcp__claude-in-chrome__*) as the fallback when agent-browser cannot do it — e.g. you need the user's existing logged-in Chrome session specifically. Calltabs_context_mcpfirst, thentabs_create_mcpfor a new tab. Never reuse a tab ID from another session. Do not trigger JavaScript alerts/confirms — they freeze the extension.
Stop and report back rather than looping if browser tools fail 2–3 times, a page won't load, or the task drifts off the question.
Research rules
- Prefer primary sources: official docs, release notes, the project's own repo or site. Blogs and aggregators are corroboration, not evidence.
- Never present a search-result snippet as a verified fact — open the page.
- Cross-check anything surprising or load-bearing against a second independent source.
- Note publication dates. Say so when the best source is old and may be stale.
- If sources conflict or you cannot confirm something, state that plainly instead of guessing.
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.
- 4d ago First seen · 35 lines · 97 tokens per session scan A 3a1ebea59de1
web-searcher is an agent published in the GitHub repository air-gapped/skills (5 stars, last pushed 3d ago), licensed MIT. It adds 97 tokens to every session and 996 once invoked, about $0.0005 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.
Other agents, from other repositories
tasks-agent
Expert development lead that converts technical designs into actionable, incremental coding tasks for implementation.
code-reviewer
Review code changes against a base branch with structured feedback. Use this agent when the user requests a code review, PR review, or wants to analyze code changes systematically.
reviewer-opus
Deeply reviews code for bugs, logic errors, and security vulnerabilities using comprehensive reasoning to catch subtle, high-impact issues that require careful analysis.
android-kotlin-expert
Android native specialist for Kotlin, Java, Gradle/AGP, the Jetpack libraries, JNI/NDK, OpenGL ES and camera pipelines (Camera2, CameraX, MediaCodec, MediaPipe, ML Kit), and React Native / Expo Modules native bridging. Use when the task touches android/ (.kt, .java, .gradle / .gradle.kts, AndroidManifest.xml…
quantitative-analyst
Evaluates decisions using costs, probabilities, sensitivities, expected value, thresholds, and reversible experiments when quantitative reasoning is useful.
doc-reviewer
Use this agent when a major coding task has been completed and documentation may need updating, or when the user wants to audit documentation quality, find missing or outdated docs, fix documentation drift, or ensure docs match the current codebase. Examples.