raffle-winner-picker
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for ingesting and processing untrusted external data from sources like CSV files, Excel spreadsheets, and Google Sheets. This creates a surface where a malicious actor could insert instructions into a raffle entry (e.g., in a name or comment field) to hijack the agent's behavior when it reads the data.
- Ingestion points: External data sources including local files (entries.csv, contest-entries.xlsx) and remote URLs (Google Sheets).
- Boundary markers: Absent; there are no instructions provided to the agent to treat the data exclusively as literals or to ignore instructions embedded within the entries.
- Capability inventory: The skill documentation describes picking winners, exporting details, and handling participant data, implying the agent will use file-reading and data-processing capabilities.
- Sanitization: Absent; the skill does not mention validating or sanitizing the content of the imported lists.
Audit Metadata