From schrute
Guides developers in choosing between Schrute skill recording and direct browser automation based on repeatability, latency needs, and API extraction goals.
How this skill is triggered — by the user, by Claude, or both
Slash command
/schrute:browser-tool-selectionThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
When a task requires browser interaction, choose the right approach based on the task characteristics.
When a task requires browser interaction, choose the right approach based on the task characteristics.
| Question | Yes → | No → |
|---|---|---|
| Will this action be repeated? | Schrute | Direct browser |
| Do I need to optimize latency? | Schrute | Either |
| Am I just looking at the page? | Direct browser | Either |
| Do I want to learn the API? | Schrute | Direct browser |
| Is this a one-time data extraction? | Direct browser | Schrute |
| Will I reuse this across sessions? | Schrute | Direct browser |
schrute_explore with the target URL to start a browser sessionschrute_record to capture an action (names the API interaction)schrute_stop to process the recording into a skillSchrute skills start at Tier 3 (browser-proxied) and can promote to Tier 1 (direct HTTP) after validation:
| Tier | Method | Latency | Promotion Criteria |
|---|---|---|---|
| 1 | Direct HTTP fetch | 1-50ms | 5 consecutive validations, low volatility |
| 2 | Cookie refresh + fetch | 5-100ms | Auth-required endpoints |
| 3 | Browser-proxied fetch | 100-500ms | Default starting tier |
| 4 | Full Playwright automation | 1-10s | Fallback for complex interactions |
You can use both approaches together:
npx claudepluginhub sheeki03/schruteExplains Schrute's self-learning browser agent workflow for recording interactions into tiered API skills, including explore-record commands, tier promotion, confirmation, and security model.
Builds reliable browser automation skills through iterative testing — runs a task, reads the trace, and improves the strategy until it passes. Supports parallel multi-task runs via sub-agents.
Provides best practices for browser automation using Playwright and Puppeteer in web testing, scraping, and AI agents. Covers selectors, auto-waits, isolation, screenshots, and anti-detection.