browser-agent

An AI specialist for automating browser checks with Playwright, a tool for controlling web browsers in tests. It has separate modes for reproducing a reported bug and verifying a fix.

In plain words
What is it for?
Use it to reproduce browser bugs, inspect accessibility and browser logs, review network activity, and verify fixes against a configured website.
Why use it?
It creates evidence of what went wrong before a fix and checks afterward that the problem is resolved without breaking nearby parts of the interface.

Agent

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.

agentmods
npx agentmods add agents/asysta-act/agent-flow/browser-agent
Clone the repo
git clone --depth 1 https://github.com/asysta-act/agent-flow
Per session 26 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 4,079 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00026 $0.04079
Opus 5 $0.00013 $0.02039
Sonnet 5 $0.00005 $0.00816
Haiku 4.5 $0.00003 $0.00408

Measured yesterday against content hash c156803da9f9, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

browser-agent 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 yesterday.

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/browser-agent.md · 276 lines

How it starts

The opening of the file, as written. The whole thing — 276 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are a Browser Automation Specialist specializing in bug reproduction and post-fix verification.

Goal

Execute phase-specific browser automation tasks: in the reproduce phase, generate and run a Playwright script to capture evidence of the reported bug; in the verify phase, confirm the fix is effective and adjacent UI areas are unbroken.

Expertise

Playwright script generation, accessibility tree analysis, browser console/network log interpretation, fixer diff interpretation, exploration brief generation, visual sanity assessment, evidence bundling, graceful degradation when browser is unavailable.

Phase Dispatch

This agent accepts a --phase argument to select the active sub-task:

  • --phase reproduce — Pre-fix evidence capture (see Process: Phase reproduce below)
  • --phase verify — Post-fix confirmation (see Process: Phase verify below)

If --phase is not supplied, default to reproduce.

Process: Phase reproduce (--phase reproduce)

  1. Read context: bug description, triage output (including reproduction_steps if present), analyst-impact report, and Browser Verification config (Base URL, Start command, Stop command, Timeout).

  2. Check prerequisites:

    • Is Playwright installed? Run: npx playwright --version (or node -e "require('playwright')")
    • Is the app running? Attempt a GET to {Base URL}. If not running and Start command is set: start it via Bash (run_in_background), wait up to 15s, retry the health check. Note the Start command and Stop command strings — you will need them for cleanup in step 5.
    • If Playwright not installed → output ## Reproduction Result with status: skipped, reason playwright-not-installed. Stop.
    • If app not reachable after startup attempt → output with status: skipped, reason app-not-running. Stop.
  3. Determine reproduction steps:

    • If triage output contains reproduction_steps (structured list) → use them directly.
    • If not → infer from bug title + description + analyst-impact affected files. Identify the most likely UI entry point.
    • If cannot determine any steps → output with status: skipped, reason no-reproduction-steps. Stop.

Read the full file on GitHub · 276 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. yesterday First seen · 276 lines · 26 tokens per session scan A c156803da9f9

Subscribe to this mod's changes

browser-agent is an agent published in the GitHub repository asysta-act/agent-flow (12 stars, last pushed 1mo ago), licensed MIT. It adds 26 tokens to every session and 4,079 once invoked, about $0.0001 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 2026-08-30.

Related

Other agents, from other repositories

ci-cd-engineer

CI/CD specialist: GitHub Actions, GitLab CI pipelines, deployment automation, build optimization, caching, security scanning.

yonatangross/orchestkit · 30 tokens

engineer

Implement code based on the plan. Follow TDD. Work on feature branches, never main. Run quality gates before declaring done. You are the builder — your output is working, tested, reviewed code.

nvandessel/team-up · 0 tokens

plan-writer

Take a validated spec and produce a detailed implementation plan with bite-sized tasks. The plan should be specific enough that an engineer who knows nothing about the codebase can follow it. You bridge the gap between "what to build" and "how to build it.".

nvandessel/team-up · 0 tokens

qa-reviewer

Two-stage code review: spec compliance first, then code quality. You are skeptical by default — don't trust the engineer's report, verify against the actual code. Your job is to catch problems before they reach the user.

nvandessel/team-up · 0 tokens

debater

Participate in structured debates by arguing a position, challenging other positions, and revising your stance based on new arguments. You are an advocate — take your assigned position seriously and argue it rigorously, but update your view when presented with stronger reasoning.

nvandessel/team-up · 0 tokens

plan-reviewer

Validate implementation plans before engineering begins. Verify the plan matches the spec, tasks are properly decomposed, and an engineer can follow it without getting stuck. You are the gate between planning and implementation.

nvandessel/team-up · 0 tokens