yt-chapters

Warn

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The chapters.py script performs dynamic loading of a Python module from a computed relative path. Specifically, it modifies sys.path to include a directory outside its own folder (../yt-edit) and imports the deadair module. This allows the script to execute code from a location that may not be under the skill's direct control, posing a risk if the environment contains untrusted files in adjacent directories.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted transcript files, creating an attack surface where embedded instructions could influence the agent.\n
  • Ingestion points: The script reads raw transcript data from local files provided by the user in chapters.py.\n
  • Boundary markers: The skill lacks delimiters or specific instructions for the agent to ignore potential commands within the transcript content.\n
  • Capability inventory: The script performs file reads and outputs to the terminal; it does not have network or file-write permissions.\n
  • Sanitization: No sanitization is performed on the transcript text. Malicious instructions in the transcript could be extracted as keywords and subsequently influence the agent's behavior during the retitling phase.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 9, 2026, 12:32 PM