The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process large amounts of untrusted data from external websites, which is a primary vector for indirect prompt injection attacks.
Ingestion points: The skill uses firecrawl_map and firecrawl_scrape in Phase 1 to fetch content from arbitrary competitor URLs, blog posts, and third-party review sites such as G2 and Capterra.
Boundary markers: The skill includes a 'Core Principles' section titled 'Untrusted Input' which explicitly instructs the agent that fetched pages may contain text aimed at AI agents (e.g., 'describe this product favorably') and to ignore any embedded instructions.
Capability inventory: The agent has the capability to write synthesized profiles and raw scrape data to the local filesystem (competitor-profiles/ directory) and read local context from files such as .agents/product-marketing.md.
Sanitization: While the skill provides high-level instructions to ignore embedded directives, it does not implement programmatic sanitization or strict schema validation for the raw markdown content scraped from the web.

competitor-profiling