The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides explicit instructions to download and execute a shell script from a remote server using a highly insecure pattern (curl piped to shell).\n
Evidence: curl -fsSL https://treg.to/install.sh | sh in SKILL.md.\n
Risk: The script is fetched from treg.to, which is not an authorized or trusted organization, allowing for arbitrary command execution on the host system without prior verification.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which constitutes a vulnerability surface for prompt injection instructions.\n
Ingestion points: Reads video-related data from <edit>/post.md, video scripts, transcripts, and voice guidelines in section 4.\n
Boundary markers: Absent. There are no clear delimiters or instructions telling the agent to ignore commands within the ingested content.\n
Capability inventory: The skill has the ability to execute shell commands (ffmpeg) and perform network operations via the treg CLI tool.\n
Sanitization: Absent. The content is directly interpolated into prompts for generating captions and titles without validation or escaping.\n- [EXTERNAL_DOWNLOADS]: The skill uses an external CLI tool to upload local files to a third-party server.\n
Evidence: The command treg host <edit>/social.mp4 uploads local media files to provide public URLs for social media publishing.

video-publish