The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes untrusted data from the internet.
Ingestion points: Untrusted data enters the agent context through search results and scraped web content from 10–15 external sources (Step 3 and Step 4).
Boundary markers: The instructions do not define explicit boundary markers or "ignore embedded instructions" warnings for the agent when it processes or summarizes the external content.
Capability inventory: The skill utilizes shell execution (bash), external search APIs (Tavily, Firecrawl), and file system write operations (Step 3, 4, 5, 6, 7).
Sanitization: No explicit sanitization, validation, or filtering of the retrieved web content is performed before it is used for summarization and synthesis.
[COMMAND_EXECUTION]: The skill relies on bash command execution to perform its primary functions. This includes running the firecrawl CLI for searching and scraping (Step 3, 4) and executing a local shell script for report verification (source "$VERIFY_LIB" in Step 7). While these are functional requirements, they represent a significant capability tier.

research