The Agent Skills Directory

[DATA_EXFILTRATION]: The skill transmits sensitive information, including person/organization names and full document files (PDFs/Images of IDs or company extracts), to an external API endpoint (https://www.stipple.sh/v1/adverse-media). While this is the primary purpose of the skill, it represents a data exposure surface for sensitive user identifiers and documentation.
[COMMAND_EXECUTION]: The instructions explicitly direct the agent to execute shell commands using curl to interact with the Stipple REST API and suggest the use of an MCP client connected to an external URL. These commands involve reading local files and transmitting them over the network.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (PDF/Image files and user-provided names) which can contain malicious instructions designed to influence the agent's interpretation of screening hits.
Ingestion points: Target names and uploaded files (company-extract.pdf) provided by users or external sources (SKILL.md, Step 1 & 2).
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially adversarial content within the ingested documents.
Capability inventory: The skill possesses network transmission capabilities (curl POST) and file system read access (-F "file=@...").
Sanitization: There is no evidence of sanitization or validation of the input data before it is processed or sent to the external API.

screen-adverse-media