The Agent Skills Directory

[SAFE]: The skill consists entirely of instructional markdown files and reference documents aimed at refining transcriptions. It does not perform any network operations, access sensitive system file paths, or include hardcoded credentials.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an ingestion surface as it processes user-provided transcription text and files (.txt, .srt, .vtt). While this is a theoretical attack surface, the instructions include specific constraints that mitigate this risk. The agent is explicitly commanded to act only as a text polisher, forbidden from adding content not present in the source, and instructed to preserve the original meaning, which prevents it from being easily diverted by instructions embedded within the transcript data.

transcript-polisher