The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it ingests untrusted data from external competitor websites using the Firecrawl tool. This risk is effectively mitigated by explicit instructions in the 'Untrusted Input' section of SKILL.md, which directs the agent to treat fetched pages as data to analyze rather than instructions to follow and to ignore any embedded directives intended for AI agents.
[DATA_EXPOSURE]: The skill reads local project configuration files, specifically .agents/product-marketing.md, to gather context about the user's product. This is a standard operational pattern for AI agents to provide relevant output and does not involve access to sensitive system credentials or private keys.
[COMMAND_EXECUTION]: The skill utilizes external tools for mapping and scraping websites (Firecrawl) and retrieving SEO metrics (DataForSEO). These operations are well-defined within the skill's primary purpose of competitive analysis and are performed using established tool schemas rather than arbitrary command-line execution.

competitor-profiling