Stagehand agent to extract screenshots of Linkedin banner images

[aq17]

Somewhat niche, somewhat interesting use case built for another customer demo.

Imagining this as an example of how to leverage Stagehand to download random media assets on websites where a user would typically "right-click and press save image" (OS-level commands) that Stagehand can't mimic with acts, but can instead extract the image links, navigate to the link, and ultimately take a screenshot of it instead

---

Note

Low Risk
Low risk because this PR only adds new standalone Python/TypeScript example projects and supporting docs/config, without modifying existing application logic. Main risk is minor dependency/lockfile churn and potential confusion around required env vars/API keys.

Overview
Adds new python/agent-screenshot and typescript/agent-screenshot examples that use a Stagehand CUA agent (via Browserbase with proxies/advanced stealth) to Google-search a company’s LinkedIn page, extract the banner image URL, and save a cropped screenshot to images/.

Includes project scaffolding and setup docs (README.md, .env.example, .gitignore), plus Python packaging (pyproject.toml, uv.lock) and runnable entrypoints (main.py, index.ts).

^{Written by Cursor Bugbot for commit 0922b56. This will update automatically on new commits. Configure here.}

[Kylejeong2]

TS version lgtm other than the fact that you can't use advanced stealth/proxies on a non-paid plan.

Also all of our templates are in both python/TS

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

There are 3 total unresolved issues (including 1 from previous review).

Autofix Details

Bugbot Autofix prepared fixes for both issues found in the latest run.

• ✅ Fixed: Wrong env var name breaks Stagehand client API key
  • Updated AsyncStagehand initialization to read model_api_key from GOOGLE_API_KEY so extract uses the configured key.
• ✅ Fixed: Missing prominent Scale plan requirement note per reviewer
  • Added a prominent note at the top of README stating the example is for Browserbase Scale customers due to advanced_stealth.

Or push these changes by commenting:

@cursor push bc5bd877a3

Preview (bc5bd877a3)

diff --git a/python/agent-screenshot/README.md b/python/agent-screenshot/README.md
--- a/python/agent-screenshot/README.md
+++ b/python/agent-screenshot/README.md
@@ -1,5 +1,7 @@
 # Stagehand + Browserbase: LinkedIn Banner Screenshot
 
+> **Note:** This example is for Browserbase Scale customers only because it requires `advanced_stealth`.
+
 ## AT A GLANCE
 
 - Goal: Given a company name, find its LinkedIn page via Google and extract a screenshot of its banner image.

diff --git a/python/agent-screenshot/main.py b/python/agent-screenshot/main.py
--- a/python/agent-screenshot/main.py
+++ b/python/agent-screenshot/main.py
@@ -85,7 +85,7 @@
     client = AsyncStagehand(
         browserbase_api_key=os.environ.get("BROWSERBASE_API_KEY"),
         browserbase_project_id=os.environ.get("BROWSERBASE_PROJECT_ID"),
-        model_api_key=os.environ.get("GOOGLE_GENERATIVE_AI_API_KEY"),
+        model_api_key=os.environ.get("GOOGLE_API_KEY"),
     )
 
     # ---------------------------------------------------------------------------

_{This Bugbot Autofix run was free. To enable autofix for future PRs, go to the Cursor dashboard.}

[cursor]

Wrong env var name breaks Stagehand client API key

High Severity

The AsyncStagehand client reads model_api_key from GOOGLE_GENERATIVE_AI_API_KEY, but the .env.example, README, error messages, and even the comment on line 83–84 all reference GOOGLE_API_KEY. The agent config on line 169 also correctly uses GOOGLE_API_KEY. This mismatch means the client-level API key will always be None, causing the extract() call to fail since it depends on the client's model_api_key for the google/gemini-2.5-flash model.

Additional Locations (1)

• python/agent-screenshot/.env.example#L9-L10

 

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

There are 4 total unresolved issues (including 3 from previous reviews).

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}