The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions include a command to fetch a shell script from a remote URL (https://treg.to/install.sh) and pipe it directly to a shell for execution (curl -fsSL ... | sh). This allows for arbitrary code execution from a domain that is not recognized as a trusted vendor or well-known service.
[EXTERNAL_DOWNLOADS]: The skill references and fetches instructions and scripts from external repositories, including raw.githubusercontent.com/agentara/skills and the author's own repository raw.githubusercontent.com/superdesigndev/treg, creating dependencies on remote content that can be modified.
[DYNAMIC_EXECUTION]: The workflow relies on the execution of external Python scripts (seedance_treg.py and caption_burn.py) which are not bundled with the primary skill file and must be loaded from external sources at runtime.
[INDIRECT_PROMPT_INJECTION]: 1. Ingestion points: The skill is designed to ingest large amounts of untrusted data, including TikTok and Instagram search results and video transcripts, into the agent's context. 2. Boundary markers: No delimiters or protective instructions are used to separate user data from instructions when processing fetched transcripts. 3. Capability inventory: The agent possesses capabilities for network calls via treg and file/process manipulation via ffmpeg and Python. 4. Sanitization: There is no evidence of sanitization or validation of the fetched social media content.
[COMMAND_EXECUTION]: The skill uses various command-line tools like ffmpeg and the treg CLI to perform its core functions, involving shell command construction that could be influenced by external inputs.

make-ugc