The Agent Skills Directory

[Indirect Prompt Injection] (MEDIUM): The skill's core functionality involves ingesting untrusted data from the web and processing it to generate a local file.
Ingestion points: Step 2 of SKILL.md requires the agent to search for trends, industry reports, and trending conversations on the web. This content is attacker-controllable.
Boundary markers: Absent. The instructions do not define delimiters or provide warnings to the agent to ignore instructions embedded within the research material.
Capability inventory: The skill possesses web browsing capabilities and the ability to write files to the local filesystem (Step 4: Save the File).
Sanitization: Absent. There is no requirement to sanitize or validate the content retrieved from the web before including it in the output file.
[Data Exposure & Exfiltration] (LOW): The skill performs network operations to browse the web. While required for research, it involves interacting with non-whitelisted external domains, which is a prerequisite for exfiltration.
[COMMAND_EXECUTION] (LOW): The skill automates the writing of a .md file to the local filesystem. If the research results contain malicious payloads (e.g., terminal escape sequences or malicious markdown), opening the resulting file could pose a minor risk to the user environment.

topic-research