The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the user to clone an external third-party repository (https://github.com/owenzlz/EgoHOS.git) and execute shell and Python scripts (bash scripts/download_models.sh, python scripts/download_pretrained_models.py) contained within. Running unverified code from an external source is a high-risk activity that can lead to unauthorized access or system compromise.
[EXTERNAL_DOWNLOADS]: The skill fetches code and data from a non-verified GitHub repository (https://github.com/owenzlz/EgoHOS.git) that is not part of any trusted organization.
[METADATA_POISONING]: The skill's metadata identifies the author as 'K-Dense Inc.', which contradicts the platform-provided author 'wu-yc'. This discrepancy is misleading regarding the skill's origin.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted media files (videos and images) and possesses high-privilege capabilities (file writing, shell execution), creating a vulnerability to indirect prompt injection.
Ingestion points: External files processed via cv2.VideoCapture and cv2.imread in SKILL.md.
Boundary markers: Absent. No delimiters or instructions are used to isolate untrusted data.
Capability inventory: The skill is capable of executing shell commands (git clone, bash, python) and writing to the file system (cv2.VideoWriter, cv2.imwrite).
Sanitization: Absent. No verification or sanitization of the input content is described.

egohos-segmentation