Skip to content
Back to search

GeminiCLI_ComputerUse_Extension

A Google Gemini-CLI extension than enables Gemini Computer Use from the CLI

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/automateyournetwork--geminicli-computeruse-extension/badge.svg)](/listings/automateyournetwork--geminicli-computeruse-extension)

HTML

<a href="/listings/automateyournetwork--geminicli-computeruse-extension"><img src="/api/listings/automateyournetwork--geminicli-computeruse-extension/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 225 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # GeminiCLI_ComputerUse_Extension
  2. A Google Gemini-CLI extension than enables Gemini Computer Use from the CLI
  3. ## Installation
  4. gemini extensions install https://github.com/automateyournetwork/GeminiCLI_ComputerUse_Extension.git
  5. ### Example Gemini-CLI prompt to VISIBLY browse a web article and produce a Markdown report
  6. Use ComputerUse MCP headfully to VISIBLY browse like a human and produce a final Markdown report.
  7. Goals:
  8. - Read the article at the given URL.
  9. - Scroll like a human (visible motion), taking a snapshot after EACH scroll step.
  10. - Visit all in-article links (avoid mailto/tel/#). Prefer opening in NEW TABS that do not overtake the main article tab. If tab tools are unavailable, open links sequentially and return to the article each time.
  11. - Conclude with a comprehensive Markdown summary (RAFT, retrieval-augmented fine-tuning, key takeaways), citing the pages you visited.
  12. 1) initialize_browser(
  13. url="https://www.automateyournetwork.ca/pyats/augmenting-network-engineering-with-raft/",
  14. width=1920, height=1080, headless=false)
  15. 2) Smoothly scroll the article TOP→BOTTOM with visible pauses. After EACH step, take a snapshot:
  16. for y in [200, 400, 600, 800, 1000]:
  17. - execute_action("scroll_to_percent", {"y": y})
  18. - capture_state(f"scroll_{y}")
  19. 3) Return to TOP visibly and snapshot:
  20. - execute_action("scroll_to_percent", {"y": 0})
  21. - capture_state("back_to_top")
  22. 4) Harvest in-article links (de-dupe, skip mailto/tel/#):
  23. - execute_action("execute_javascript", {
  24. "code": "(() => { \
  25. const scope = document.querySelector('article, main, .post-content, .entry-content') || document.body; \
  26. const links = Array.from(scope.querySelectorAll('a[href]')); \
  27. const hrefs = links.map(a => a.href.trim()) \
  28. .filter(h => h && !h.startsWith('mailto:') && !h.startsWith('tel:') && !h.includes('#')); \
  29. return Array.from(new Set(hrefs)); \
  30. })();"
  31. })
  32. 5) Visit each harvested link (limit to 8 to stay readable):
  33. Preferred (if these tools exist): open_new_tab, list_tabs, switch_to_tab
  34. - For each <link> (1-based index i):
  35. - If open_new_tab exists:
  36. - open_new_tab(url="<link>", focus=true)
  37. - capture_state(f"link{i}_load") # snapshot immediately after load
  38. - For y in [200, 400, 600, 800, 1000]:
  39. execute_action("scroll_to_percent", {"y": y})
  40. capture_state(f"link{i}_scroll_{y}") # snapshot each scroll step
  41. - switch_to_tab(0) # return to main article tab
  42. - Else (fallback when tab tools absent):
  43. - execute_action("open_web_browser", {"url": "<link>"})
  44. - capture_state(f"link{i}_load")
  45. - For y in [200, 400, 600, 800, 1000]:
  46. execute_action("scroll_to_percent", {"y": y})
  47. capture_state(f"link{i}_scroll_{y}")
  48. - execute_action("open_web_browser", {"url": "https://www.automateyournetwork.ca/pyats/augmenting-network-engineering-with-raft/"})
  49. - capture_state(f"link{i}_return") # confirm we’re back on the main article
  50. 6) After links, return to the main article (if not already there) and capture_state("final_overview").
  51. 7) Produce a comprehensive Markdown report (no extra screenshots in the report; just text). Structure:
  52. # RAFT & Retrieval-Augmented Fine-Tuning — Field Notes
  53. - **Primary article:** title + URL
  54. - **Other pages visited:** bullet list of titles + URLs
  55. - **What RAFT is:** 3–6 bullets (your own words)
  56. - **How RAFT differs from standard fine-tuning:** bullets
  57. - **Retrieval-Augmented Fine-Tuning pipeline:** concise steps (data prep, retrieval store, adapters/LoRA/full FT, eval)
  58. - **Cloud vs Local comparison (from the two-part series):** capabilities, privacy, cost, constraints
  59. - **Implementation notes spotted in pages:** tools, commands, pitfalls
  60. - **Key takeaways:** 5–8 bullets
  61. - **References (visited):** list of URLs
  62. Important:
  63. - Keep actions human-visible (no instant jumps).
  64. - If a page won’t load, skip it and continue.
  65. - Take a small pause between scroll steps so motion is obvious.
  66. ### Example Searching Wikipedia for "Computer Networking" and producing a Markdown report
  67. #### Wikipedia Exploration: Computer Networks
  68. Headless=false for visible browsing.
  69. You will systematically explore Wikipedia (and optionally Google) articles about computer networks using the ComputerUse MCP tools. Use selectors (not coordinates), add pauses so the typing is obvious, and take screenshots after each major step.
  70. Phase 1: Initialize & Search (Wikipedia)
  71. Initialize headful browser
  72. initialize_browser(url="https://www.wikipedia.org", width=1920, height=1080, headless=false)
  73. capture_state("wikipedia_home")
  74. pause(800)
  75. Search for “Computer Network” with visible typing
  76. fill_selector('input[name="search"]', 'Computer Network', true, true)
  77. pause(800) # let navigation/rendering be seen
  78. capture_state("wiki_search_results_or_article")
  79. Note: Wikipedia may jump straight to the article. Continue from whatever page loads.
  80. Phase 2: Explore the page/results with smooth pacing
  81. Scroll (top → bottom) with pauses
  82. for y in [200, 400, 600, 800, 1000]:
  83. execute_action("scroll_to_percent", {"y": y})
  84. pause(700)
  85. capture_state(f"wiki_scroll_{y}")
  86. Return to top
  87. execute_action("scroll_to_percent", {"y": 0})
  88. pause(500)
  89. capture_state("wiki_top")
  90. Phase 3: Extract & Visit Links (up to 8)
  91. Extract article links on current page
  92. execute_action("execute_javascript", {
  93. "code": "(() => {\
  94. const links = Array.from(document.querySelectorAll('a[href*=\"/wiki/\"]'))\
  95. .map(a => a.href)\
  96. .filter(h => h.includes('wikipedia.org/wiki/') && !h.includes('#'))\
  97. .filter((v, i, a) => a.indexOf(v) === i);\
  98. return links.slice(0, 8);\
  99. })()"
  100. })
  101. pause(600)
  102. Visit each link in sequence
  103. For each link i:
  104. execute_action("open_web_browser", {"url": "{{link_i}}"})
  105. pause(900)
  106. capture_state(f"article_{i}_loaded")
  107. for y in [0, 250, 500, 750, 1000]:
  108. execute_action("scroll_to_percent", {"y": y})
  109. pause(650)
  110. capture_state(f"article_{i}_scroll_{y}")
  111. # Go back to the search/article hub if desired:
  112. # execute_action("open_web_browser", {"url": "<the-page-you-extracted-from>"})
  113. # pause(600)
  114. (Optional) Phase 3b: Also show a Google search (for demo effect)
  115. execute_action("open_web_browser", {"url": "https://www.google.com"})
  116. pause(700)
  117. capture_state("google_home")
  118. fill_selector('textarea[name=\"q\"]', 'Computer Network', true, true)
  119. pause(900)
  120. capture_state("google_results_top")
  121. execute_action("scroll_to_percent", {"y": 600})
  122. pause(700)
  123. capture_state("google_results_mid")
  124. Phase 4: Summary Report (after all visits)
  125. After visiting all targets, produce a Markdown summary:
  126. # Wikipedia Exploration Report: Computer Networks
  127. ## Search Query
  128. - **Primary Query:** Computer Network
  129. - **Search URL:** [Insert the URL where the search happened]
  130. ## Articles Visited (up to 8)
  131. 1. [Title 1](URL) — 1–2 sentence summary
  132. 2. [Title 2](URL) — 1–2 sentence summary
  133. ...
  134. ## Key Learnings
  135. - ...
  136. - ...
  137. ## Major Subtopics
  138. - **Protocols & Standards:** ...
  139. - **Topology & Architecture:** ...
  140. - **History & Development:** ...
  141. - **Technologies & Components:** ...
  142. - **Related Disciplines:** ...
  143. ## All URLs Visited
  144. 1. URL
  145. 2. URL
  146. ...
  147. ## Environment Variables
  148. 🧩 Recommended ComputerUse MCP Environment Variables
  149. Variable Purpose Recommended for Demo Example
  150. CU_HEADFUL Launch the browser with a visible window. ✅ Yes export CU_HEADFUL=1
  151. CU_SLOW_MO Milliseconds of delay between Playwright actions (move, click, type). ✅ Yes export CU_SLOW_MO=700
  152. CU_SHOW_CURSOR Display cyan “cursor ring” overlay to visualize movement. ✅ Yes export CU_SHOW_CURSOR=true
  153. CU_NO_SANDBOX Disable Chromium sandbox if Playwright complains (needed in some Docker/macOS setups). optional export CU_NO_SANDBOX=1
  154. CU_BROWSER Force a specific browser (chromium, firefox, webkit) if you installed all. optional export CU_BROWSER=chromium
  155. CU_DEVICE_SCALE Override Retina scaling (use 2 on macOS for pixel-accurate clicks). optional export CU_DEVICE_SCALE=2
  156. 🧠 Typical macOS Demo Setup
  157. export CU_HEADFUL=1
  158. export CU_SLOW_MO=800
  159. export CU_SHOW_CURSOR=true
  160. export CU_DEVICE_SCALE=2
  161. export CU_NO_SANDBOX=0

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.