The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill includes an input parameter customMapFunction described as a 'JavaScript function to transform each output object'. This provides a surface for arbitrary code execution within the scraping environment if malicious code is supplied by a user or third-party source.
[COMMAND_EXECUTION]: The instructions rely on executing shell commands, specifically node scripts/run_actor.js and curl. While these are used for the skill's primary function, they introduce risks of command injection if user-supplied searchTerms or queries are interpolated into these shell strings without rigorous sanitization.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data in the form of searchTerms and startUrls. If this data contains malicious instructions and the agent processes it without proper boundary markers or sanitization before passing it to subsequent tools or display, it could influence the agent's behavior.
Ingestion points: searchTerms and startUrls parameters in SKILL.md.
Boundary markers: None explicitly defined in the workflow for separating data from instructions.
Capability inventory: The skill uses node subprocess calls and curl network operations.
Sanitization: No sanitization logic is described for the input parameters before they are used in commands.
[CREDENTIALS_UNSAFE]: The skill requires and utilizes APIFY_TOKEN. The provided examples show the token being used as a environment variable in shell commands (token=$APIFY_TOKEN). While using environment variables is standard practice, passing secrets via command-line arguments or URL parameters can lead to exposure in process lists, shell history, or server logs.

scraping-google-search-results