Security audit
Security checks for vulnerabilities and agentic risk
This skill is a small, disclosed Browser Use cloud API wrapper, but users should treat tasks and browser artifacts as data sent to a third party.
Install only if you intend to use Browser Use's cloud service. Do not send secrets, credentials, regulated data, private internal URLs, or irreversible account actions unless you have reviewed Browser Use's policies and are comfortable with remote processing of task text, visited pages, outputs, steps, and screenshots.
Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.
fi
# Submit task
RESPONSE=$(curl -s -X POST "$API_URL/tasks" \
-H "X-Browser-Use-API-Key: $BROWSER_USE_API_KEY" \
-H "Content-Type: application/json" \
-d "{\"task\": $(echo "$TASK" | python3 -c 'import json,sys; print(json.dumps(sys.stdin.read().strip()))')}")
Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.
# Poll for completion
ELAPSED=0
while [[ $ELAPSED -lt $MAX_WAIT ]]; do
RESULT=$(curl -s "$API_URL/tasks/$TASK_ID" \
-H "X-Browser-Use-API-Key: $BROWSER_USE_API_KEY")
STATUS=$(echo "$RESULT" | python3 -c "import sys,json; print(json.load(sys.stdin).get('status',''))" 2>/dev/null || echo "")
The skill documents shell-based execution (curl and helper scripts) but does not declare an explicit tool scope such as allowed tools or permissions. That can lead to unintended shell access during invocation and weakens policy enforcement and review, especially for a skill that transmits user tasks to an external service.
The trigger phrases are broad (scrape website, automate web task, cloud browser) and may cause the skill to activate in situations where a safer local or lower-privilege option would be more appropriate. Because this skill sends tasks to a third-party browser automation service, accidental invocation can expose user instructions and browsing targets without clear intent.
The skill describes sending plain-English browsing tasks and retrieving outputs, steps, and screenshots from a cloud service, but it does not warn users that their instructions, visited pages, and interaction artifacts may be transmitted to and stored by a third party. This creates a meaningful privacy and data-handling risk, especially if users provide credentials, sensitive URLs, or internal data.
This example performs a POST request to a third-party API and includes a user-supplied task payload for remote execution. In this skill's context, that means browsing instructions and potentially sensitive targets or content are transmitted off-platform, with additional risk from API-key use and remote processing.
# Submit task
curl -s -X POST https://api.browser-use.com/api/v2/tasks \
-H "X-Browser-Use-API-Key: $BROWSER_USE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"task": "Go to example.com and extract the main heading"}'
This example performs a POST request to a third-party API and includes a user-supplied task payload for remote execution. In this skill's context, that means browsing instructions and potentially sensitive targets or content are transmitted off-platform, with additional risk from API-key use and remote processing.
# Submit task
curl -s -X POST https://api.browser-use.com/api/v2/tasks \
-H "X-Browser-Use-API-Key: $BROWSER_USE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"task": "Go to example.com and extract the main heading"}'
Polling task status from the external service retrieves remote execution results and may expose outputs, step traces, and screenshots generated from prior browsing activity. While expected for the product, it still represents ongoing interaction with a third-party service that may surface sensitive content to systems outside the local environment.
-d '{"task": "Go to example.com and extract the main heading"}'
curl -s "https://api.browser-use.com/api/v2/tasks/TASK_ID"
-H "X-Browser-Use-API-Key: $BROWSER_USE_API_KEY"
The balance-check example transmits the API key to the external service. Although routine, it reinforces that the skill depends on shell commands making authenticated outbound requests and should therefore clearly communicate credential handling and restrict where such commands can run.
~$0.01-0.05 per task depending on complexity. Check balance:
curl -s https://api.browser-use.com/api/v2/credits \
-H "X-Browser-Use-API-Key: $BROWSER_USE_API_KEY"
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
set -euo pipefail
API_URL="https://api.browser-use.com/api/v2"
POLL_INTERVAL=5
MAX_WAIT=300
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
set -euo pipefail
API_URL="https://api.browser-use.com/api/v2"
POLL_INTERVAL=5
MAX_WAIT=300
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
set -euo pipefail
API_URL="https://api.browser-use.com/api/v2"
POLL_INTERVAL=5
MAX_WAIT=300
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
set -euo pipefail
API_URL="https://api.browser-use.com/api/v2"
POLL_INTERVAL=5
MAX_WAIT=300
The script sends the user-provided task string to a third-party cloud API along with an API key. In this skill's context, that means potentially sensitive prompts, URLs, credentials, or scraped data may leave the local environment and be processed remotely, which is a real data-exposure risk even though it is core to the tool's stated purpose.
fi
# Submit task
RESPONSE=$(curl -s -X POST "$API_URL/tasks" \
-H "X-Browser-Use-API-Key: $BROWSER_USE_API_KEY" \
-H "Content-Type: application/json" \
-d "{\"task\": $(echo "$TASK" | python3 -c 'import json,sys; print(json.dumps(sys.stdin.read().strip()))')}")
No suspicious patterns detected.