wjs-explaining-tv-slang
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is exclusively focused on educational linguistic analysis of television dialogue. The logic provided in the instructions is transparent and matches the stated purpose without any signs of malicious intent or hidden code execution.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted data from external sources such as subtitle (SRT) files and screenshots. While this creates a vulnerability surface for indirect prompt injection, the agent's capabilities are restricted to linguistic explanation and writing to a specific local file, which limits the potential impact of such an attack.
- [DATA_EXFILTRATION]: The skill maintains a local vocabulary notebook by appending data to
~/code/tv-slang/notebook.md. This is a functional feature intended for user progress tracking and is confined to a non-sensitive directory, posing no risk of sensitive credential harvesting or unauthorized external transmission.
Audit Metadata