yt-package
Pass
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill instructions direct the agent to read from
~/.youtube-agent-skill/voice.md. Accessing hidden directories and files in the user's home folder is a sensitive operation that can be used to gather user information. - [DATA_EXPOSURE_AND_EXFILTRATION]: The
title.pyscript is designed to read a file path provided as a command-line argument usingopen(a[0]).read().splitlines(). If the agent is instructed to provide a sensitive file path (such as.envor SSH keys) to this script, the script will read the file and output its content as "titles", exposing the data to the agent's context. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from multiple sources which could contain malicious instructions.
- Ingestion points: The skill reads
youtube-voice.mdfrom the project directory,~/.youtube-agent-skill/voice.mdfrom the home directory, and any file path provided as an argument to thetitle.pyscript. - Boundary markers: No explicit delimiters or boundary markers are used to wrap the ingested content to prevent the agent from following embedded instructions.
- Capability inventory: The skill possesses file read capabilities via the agent's tools and the ability to execute the local Python script via subprocess.
- Sanitization: The
title.pyscript performs no sanitization or path validation on the input file, and the instructions do not include steps to sanitize data before interpolation.
Audit Metadata