AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
SKILL verified MIT Self-run

Benchmark Browsing Skills

skill-browsing-skills-browsing-skills-benchmark-browsing-skills · by browsing-skills

Use when benchmarking browsing-skills actions, comparing a maintained site skill against a no-skill browser agent, measuring OpenAI API token usage, wall time, API calls, browser/tool calls, and result quality. Trigger whenever the user asks to benchmark a browsing skill, compare with/without skill, measure token or time savings, or update benchmark tables in this repo.

↗ Repository
— No reviews yet
0 installs
39 views
0.0% view→install

Install

$ agentstack add skill-browsing-skills-browsing-skills-benchmark-browsing-skills

✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • ✓ Prompt-injection patterns
  • ✓ Secret / credential exfiltration
  • ✓ Dangerous shell & filesystem operations
  • ✓ Untrusted network calls
  • ✓ Known-malicious package signatures

What it can access

  • ✓ Network access No
  • ✓ Filesystem access No
  • ✓ Shell / process execution No
  • ✓ Environment & secrets No
  • ✓ Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README, it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/skill-browsing-skills-browsing-skills-benchmark-browsing-skills)

Reliability & compatibility

✓ Security review passed
0 installs to date
— no reviews yet
● 3mo ago

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Benchmark Browsing Skills? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Benchmark Browsing Skills

Use this skill to run and document fair benchmarks for actions in this repo.

The benchmark compares two agents on the same live browser task:

  • With skill: loads the site action index plus only the action-facing parts of one action reference, then runs the maintained action code.
  • Without skill: receives browser access but no skill/reference, inspects the live page, derives extraction logic, and iterates until success or timeout.

For official numbers, prefer OpenAI API-separated runs over Codex subagents. Subagents are useful for quick qualitative comparisons, but API runs give explicit token accounting and cleaner cost comparison.

Use the same browser layer and session state for both branches. For actions that depend on login, personalization, region, cart state, or existing cookies, prefer Chrome Bridge /run-action against the user's loaded Chrome session. Use Playwright only for public, unauthenticated actions or when you intentionally create an equivalent browser profile for both branches.

Workflow

  1. Identify the website, action, target URL/query, browser layer, expected output fields, session requirements, and success criteria.
  2. Read only the matching site SKILL.md and action reference from skills//references/.md; exclude benchmark history and other non-action prose from the measured with-skill prompt.
  3. Run the with-skill API pass with the skill index/reference available.
  4. Run the without-skill API pass without skill files; provide only browser tools and live page observations.
  5. Capture tokens, API calls, wall time, browser calls, and result quality for both passes.
  6. Record results in:
  • BENCHMARKS.md
  • the site skills//SKILL.md benchmark table
  • the action reference ## Benchmark section

Read [references/api-benchmark-protocol.md](references/api-benchmark-protocol.md) before designing or running a benchmark.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Write a review

Versions

  • v0.1.0 Imported from the upstream source.

More in AI & ML