The Agent Skills Directory

[SAFE]: The skill utilizes well-known external services for data acquisition. It leverages Firecrawl for web scraping and DataForSEO for market intelligence metrics. These are standard tools for competitive analysis tasks.
[SAFE]: Data persistence best practices are implemented. The skill saves raw scraped content and JSON responses to a structured local directory (competitor-profiles/raw/) using date-stamped folders, which facilitates auditing and reduces redundant network activity.
[SAFE]: The skill follows a least-privilege approach to information gathering, checking for existing project context (e.g., .claude/product-marketing.md) before requesting user input.
[PROMPT_INJECTION]: The skill processes untrusted web content as its primary function, which inherently involves an indirect prompt injection surface. This is considered safe given it is the primary purpose of the skill and the risks are limited to the generated reports.
Ingestion points: Data from firecrawl_scrape and firecrawl_search (SKILL.md, Phase 1) enters the context from arbitrary external URLs.
Boundary markers: The synthesis phase (Phase 3) uses structured markdown templates to organize the data.
Capability inventory: The skill's capabilities are focused on reading local context and writing markdown profiles to the competitor-profiles/ directory.
Sanitization: The skill relies on the scraping tool to convert HTML to clean markdown, reducing the risk of malicious scripts.

competitor-profiling