The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external websites, creating a potential surface for indirect prompt injection attacks.
Ingestion points: Untrusted content is ingested from arbitrary URLs via page.evaluate() in scraping-browser.ts and scrape-products.ts.
Boundary markers: No specific boundary markers or instructions to ignore embedded commands are present in the scripts to protect the agent from interpreting instructions found within scraped content.
Capability inventory: The skill is configured with access to Write and Bash tools, which could be targeted by instructions embedded in external site content.
Sanitization: The example scripts do not implement sanitization or validation of the extracted data before it is returned to the agent context.
[EXTERNAL_DOWNLOADS]: Fetches standard browser automation libraries and binaries.
Evidence: The instructions include commands to install playwright via npm and the Chromium browser via npx playwright install.
[COMMAND_EXECUTION]: Performs environment setup through shell commands.
Evidence: Uses npm install and npx commands in Step 1 to prepare the local environment for scraping.

brightdata-core-workflow-a