The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external competitor websites via the Firecrawl tool suite, which naturally introduces a risk of indirect prompt injection where an attacker could place hidden instructions on a website to influence the agent.\n
Ingestion points: Site content extracted from competitor URLs via firecrawl_scrape and firecrawl_map (SKILL.md, Phase 1).\n
Boundary markers: While no technical delimiters (like XML tags) are specified, the skill includes a dedicated "Principle 5: Untrustworthy inputs" section that explicitly instructs the agent to ignore any IA-directed instructions found in HTML or review text.\n
Capability inventory: The agent has the ability to write structured markdown profiles to the local competitor-profiles/ directory and make network requests via MCP tools (Firecrawl and DataForSEO).\n
Sanitization: There is no evidence of automated content sanitization or filtering applied to the raw scraped data before it is saved or processed.\n- [DATA_EXFILTRATION]: The skill accesses local project context files, such as .agents/product-marketing.md, to tailor its analysis. It also performs network operations to fetch data from well-known services like DataForSEO. These actions are consistent with the skill's primary objective of competitive intelligence and rely on legitimate, industry-standard tools.\n- [SAFE]: No evidence of obfuscation, hardcoded credentials, persistence mechanisms, or privilege escalation was found in the skill's instructions or metadata. The execution flow follows established patterns for AI-assisted market research.

competitor-profiling