The Agent Skills Directory

[PROMPT_INJECTION]: The skill establishes a large surface for indirect prompt injection due to its primary function of scraping and reporting data from arbitrary external websites. Maliciously crafted content on target websites could attempt to influence the agent's behavior during the summarization or reporting phase.
Ingestion points: Scraped data retrieved via the export_data tool and external files downloaded using directAccess links (found in SKILL.md and references/gotchas.md).
Boundary markers: The instructions do not prescribe specific delimiters or safety warnings for the agent when reporting the scraped content back to the user.
Capability inventory: The agent has access to tools for starting/stopping tasks (start_or_stop_task), executing new tasks (execute_task), and performing shell commands via the curlTemplate mechanism.
Sanitization: No explicit content filtering or escaping requirements are defined for the processed data.
[COMMAND_EXECUTION]: The workflow for handling large result sets (50+ rows) directs the agent to execute a curlTemplate provided dynamically in the response from the export_data tool. While this is a functional requirement for retrieving data from the vendor's infrastructure (Octoparse), it involves the agent executing shell commands generated at runtime by a remote service.

octoparse-ultimate-scraper