The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to manage its output directory on the local filesystem. It executes commands such as mkdir -p to create research-specific subfolders and rm to clean up files that do not meet the quality threshold within the path F:\Dropbox\CC\选题大全\.
[EXTERNAL_DOWNLOADS]: The skill retrieves academic data from well-known and reputable sources, including top-tier finance journals, the SSRN repository, and seminar schedules from established universities such as Chicago Booth, Wharton, and MIT Sloan. This data collection is fundamental to the skill's purpose of identifying research gaps.
[PROMPT_INJECTION]: The skill processes untrusted data retrieved from external academic sources, creating an attack surface for indirect prompt injection.
Ingestion points: Data is ingested via WebSearch and WebFetch operations targeting academic journals, SSRN abstracts, and university seminar pages.
Boundary markers: The instructions for subagents lack explicit delimiters or sanitization commands to isolate potential instructions embedded in the paper titles or abstracts.
Capability inventory: The skill and its subagents have access to Write, Bash, Agent, and Skill tools, which could be leveraged if malicious instructions were successfully injected from a remote source.
Sanitization: There is no mention of filtering or sanitizing the external literature content before it is used to prompt the subagents.

econfin-idea-finder