Skip to content

cli.md

Latest commit

 

History

History
158 lines (111 loc) · 4.27 KB

cli.md

File metadata and controls

158 lines (111 loc) · 4.27 KB

CLI Reference

Basic Usage

# single task
bad run --goal "Sign up for account" --url http://localhost:3000

# test suite
bad run --cases ./cases.json

# with config file
bad run --config ./ci.config.ts --cases ./cases.json

# override model/concurrency
bad run --cases ./cases.json --model gpt-5.4 --concurrency 4

Authenticated Sessions

Save a browser session once, reuse across runs:

pnpm auth:save-state
pnpm auth:check-state ./.auth/session.json example.com
bad run --goal "Open settings" \
  --url https://app.example.com \
  --storage-state ./.auth/session.json

Run Modes

Mode Vision Screenshots Blocking Use case
fast-explore off off analytics Speed / iteration
full-evidence on every 3 turns — Release signoff

Mode presets apply defaults; explicit CLI flags override.

bad run --cases ./cases.json --mode fast-explore
bad run --cases ./cases.json --mode full-evidence

Execution Profiles

Profile Description
default Balanced defaults
stealth Headed + anti-detection args
webbench Speed benchmark (vision off, heavy blocking)
webbench-stealth Reach benchmark (stealth args, analytics-only blocking)
webvoyager Evidence benchmark (vision on)

Profiles are orthogonal to modes. Use both:

bad run --cases ./cases.json --profile webbench-stealth --mode fast-explore

Adaptive Model Routing

Route verification to a cheaper model:

bad run \
  --model gpt-5.4 \
  --model-adaptive \
  --nav-model gpt-4.1-mini \
  --cases ./cases.json

Trajectory Memory

Memory is enabled by default. Successful run trajectories are stored in .agent-memory/ and reused on subsequent runs to reduce turns and improve reliability.

# disable memory for a clean run
bad run --cases ./cases.json --no-memory

# custom memory directory
bad run --cases ./cases.json --memory-dir ./.my-memory

# with trace scoring
bad run --cases ./cases.json --trace-scoring --trace-ttl-days 30

Personas

# auto-generated from goal + URL
bad run --goal "..." --url https://... --persona auto

# named persona
bad run --goal "..." --url https://... --persona alice-blueprint-builder

Design Audit

LLM-powered design quality audit with domain-specific rubrics:

bad design-audit --url https://stripe.com
bad design-audit --url https://app.uniswap.org --profile defi
bad design-audit --url http://localhost:3000 --profile saas --pages 10

Profiles: general, saas, defi, marketing.

Token Extraction

Pure DOM extraction — no LLM calls. Captures colors, typography, spacing, components, logos, icons, videos, CSS variables, and brand assets at mobile/tablet/desktop viewports. Detects inline libraries (GSAP, Three.js, p5.js, Lottie, Swiper, etc.).

bad design-audit --url https://stripe.com --extract-tokens
bad design-audit --url https://app.example.com --extract-tokens --json

Output: tokens.json + downloaded fonts, images, videos, stylesheets, screenshots.

Site Rip

Download a full working local copy of a website. Uses Playwright network interception to capture every request/response, rewrites HTML/CSS references to local paths.

bad design-audit --url https://example.com --rip
bad design-audit --url https://example.com --rip --pages 10

Reveals hidden content (accordions, tabs, carousels), auto-scrolls for lazy-loaded assets, extracts video URLs from rendered DOM. Output is a self-contained directory that opens in a browser.

Design Compare

Side-by-side comparison of two URLs with pixel diff and structural token diff.

bad design-audit --url https://site-a.com --design-compare --compare-url https://site-b.com

Captures screenshots at mobile/tablet/desktop viewports. Interacts with the page before capture:

  • Expands accordions and <details> elements
  • Clicks all tabs in tab lists
  • Scrolls carousels
  • Opens mobile hamburger menus
  • Dismisses cookie banners and modals

Output: HTML report with side-by-side screenshots + diff overlay, JSON report with structural token differences (colors, fonts, CSS variables, spacing, brand, components).

CI Checks

pnpm lint
pnpm check:boundaries
pnpm test