The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface because it processes untrusted data from external search results.
Ingestion points: The skill ingests data from the internet via the exa-search__web_search_advanced_exa tool as part of its daily and on-demand scanning workflows.
Boundary markers: There are no explicit delimiters or specific instructions in the workflow to isolate the retrieved web snippets from the agent's logic or to explicitly mark them as untrusted.
Capability inventory: The skill interacts with the pipeline-manager tool (which likely modifies local data state) and provides automated reports to the chat interface.
Sanitization: The instructions do not describe a process for sanitizing, validating, or escaping search result content before it is used for scoring or reporting to the user.

signal-scanner