The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses a risk surface for indirect prompt injection due to its core function of retrieving and processing content from external websites.
Ingestion points: The references/market-researcher.md file specifies that the agent performs web searches and collects data from sources like DART and KOSIS.
Boundary markers: The instructions lack explicit boundary markers or 'ignore' instructions for data processed from the web.
Capability inventory: The skill includes powerful tools such as bash, read, and write for data analysis as defined in the 'Market Researcher' toolset.
Sanitization: There are no provided instructions for sanitizing or escaping the text aggregated from external sources before it is processed by the model.
[COMMAND_EXECUTION]: The skill configuration in references/market-researcher.md authorizes the use of the bash command-line tool. While this is likely intended for data parsing, it provides a functional path for potentially malicious actions if the agent is successfully compromised via indirect prompt injection.

consult-market