tough-tongue/toughtongue-skills / skills/ttai-browser-demo-builder
ttai-browser-demo-builder
VerifiedScripts deterministic browser demo steps (tools_config.tools.browser.tool_settings.steps) on an existing Tough Tongue AI Scenario, so its voice agent clicks exact XPath selectors instead of improvising. Signs the demo browser in with ttai:authenticate_browser, converts a pasted browser recording or a walked-through flow into steps, saves them with ttai:update_scenario without losing the browser profile, and repairs steps that click the wrong element. Use when the user says "record browser demo steps", "turn my recording into steps", "make my demo deterministic", "the demo clicks the wrong thing", or "log the demo browser in". Not for creating the Scenario (ttai-agent) or analyzing sessions (ttai-session-analyst).
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 8.5 KB |
| agents/openai.yaml | config | 500 B |
| references/recording-to-steps.md | markdown | 7.8 KB |
| references/selector-guide.md | markdown | 7.1 KB |
| references/steps-format.md | markdown | 5.7 KB |
| references/worked-example.md | markdown | 5.2 KB |
references/worked-example.md
View on GitHub# Worked Example — Platform Tour Demo
## Contents
- The demo
- The steps
- Why each selector looks this way
- The matching instructions
- What was left out
---
## The demo
An AI SDR demos the Tough Tongue AI web app live: library, a course, then a
session analysis. Four steps cover the scripted clicks; everything else is
narration, `goto`, and three capture milestones. Treat the selectors as the
pattern — page labels may differ in the current app.
## The steps
Merged into the `tool_settings` read from the Scenario (`initialUrl` and
`contextId` carried over unchanged — see [steps-format.md](steps-format.md)):
```json
{
"library_show_coaches": {
"description": "On the library: select the Featured tab and bring the coaching row on screen",
"actions": [
{
"description": "Featured tab on the library page",
"method": "click",
"selector": "xpath=//button[@role='tab'][contains(text(),'Featured')]",
"arguments": []
},
{
"description": "Google PM Interviewer card in the featured library grid",
"method": "scrollIntoView",
"selector": "xpath=//h3[contains(text(),'Google PM Interviewer')]",
"arguments": []
}
]
},
"library_show_tutors": {
"description": "Bring the tutors and language coaches row on screen",
"actions": [
{
"description": "Interactive Python 101 card in the featured library grid",
"method": "scrollIntoView",
"selector": "xpath=//h3[contains(text(),'Interactive Python 101')]",
"arguments": []
}
]
},
"open_course": {
"description": "On the courses page: select Featured and open the Sales Coaching course",
"actions": [
{
"description": "Featured tab on the courses page",
"method": "click",
"selector": "xpath=//button[@role='tab'][contains(text(),'Featured')]",
"arguments": []
},
{
"description": "Sales Coaching: Varied Collection course card",
"method": "click",
"selector": "xpath=//h3[contains(text(),'Sales Coaching: Varied Collection')]",
"arguments": []
}
]
},
"analyze_first_session": {
"description": "On the sessions page: open the analysis of the newest completed session",
"actions": [
{
"description": "Analyze button on the first completed session row",
"method": "click",
"selector": "xpath=//tbody/tr[contains(.,'completed')][1]//button[contains(.,'Analyze')]",
"arguments": []
}
]
}
}
```
## Why each selector looks this way
- `//button[@role='tab'][contains(text(),'Featured')]` — separate brackets, not
`[@role='tab' and normalize-space()='Featured']` (forbidden pattern 3 would
collapse it to the first button on the page).
- `//h3[contains(text(),'…')]` — targets the heading, not `div[.//h3[...]]`
(forbidden pattern 1). Clicks bubble to the card.
- `scrollIntoView` for the "scroll" steps: the visible effect is the smooth
scroll replay performs before any XPath action. `scrollTo` would only move the
heading's own (nonexistent) scrollbar.
- `//tbody/tr[contains(.,'completed')][1]//button[…]` — a step-level `[1]`, not
`(//tr[...])[1]` (forbidden pattern 2). `.` reads the row's full text in both
engines, so the nested status badge matches; `text()` would miss it on a plain
page.
- Clicking Featured first is a harmless no-op when it is already active and a
guard when the page remembered another tab.
## The matching instructions
Abridged to the browser parts:
```text
## BROWSER TOOL INSTRUCTIONS
The browser is pre-opened at https://app.toughtongueai.com/. Use `runBrowserCommand`:
- `command: "step"` + `name` — pre-recorded batch. PRIMARY command.
- `command: "goto"` + `url` — direct navigation.
- `command: "capture"` — screenshot. Only at the 3 milestones below.
- `command: "observe"` / `command: "act"` — off-script moments only.
Steps run in the background — keep talking. If a step fails you are told
which action broke: `capture`, then recover with `act`.
None of the steps take variables.
## DEMO FLOW
### Phase 1: Home
1. `capture` — MILESTONE 1: see the home page, then introduce the platform.
### Phase 2: Library
1. `goto` "https://app.toughtongueai.com/library"
2. Narrate: sales and SDR agents you can remix.
3. `step` "library_show_coaches" — narrate coaching and interview agents.
4. `step` "library_show_tutors" — narrate language coaches and tutors.
### Phase 3: Courses
1. `goto` "https://app.toughtongueai.com/course"
2. `step` "open_course" — narrate: agents bundle into courses with progress.
3. `capture` — MILESTONE 2: describe the course page.
### Phase 4: Analytics
1. `goto` "https://app.toughtongueai.com/sessions"
2. `step` "analyze_first_session" — narrate while it runs.
3. `capture` — MILESTONE 3: walk through the analysis.
```
Rhythm: every page change is a `goto`, every scripted click is a `step`,
narration covers execution, captures only where the agent must see the page.
## What was left out
- **Typing.** No `fill` actions; the demo shows forms and frames typing as the
prospect's turn.
- **Date pickers and dropdown cascades.** Only tabs, cards, and one button.
- **Variables.** Every step is fixed, and the instructions say so.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |