The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill clones a software repository from an untrusted source: https://github.com/Coyote-git/gifhorse. While the skill author is noted as openclaw, the resources are hosted under a different GitHub user account.
[REMOTE_CODE_EXECUTION]: The installation sequence includes pip install -e . within the cloned directory. This command executes the setup.py or build scripts of the downloaded repository, which constitutes execution of remote code from an unverified source.
[COMMAND_EXECUTION]: The skill makes extensive use of shell commands for setup and operation, including brew install ffmpeg-full and various sub-commands of the gifhorse CLI tool which interact with the local file system.
[PROMPT_INJECTION]: The skill processes untrusted external data by downloading subtitles from online providers for transcription.
Ingestion points: Subtitles automatically fetched from third-party providers during the transcribe and fetch-subtitles operations.
Boundary markers: None identified in the prompt templates.
Capability inventory: File system read/write, network access for subtitle fetching, and subprocess execution via gifhorse and ffmpeg.
Sanitization: No evidence of sanitization for the dialogue content extracted from external subtitles before it is processed by the agent.

gifhorse