The Agent Skills Directory

[PERSISTENCE_MECHANISMS]: The skill instructs the AI agent to persistently modify the user's shell configuration files (such as ~/.zshrc and ~/.bashrc) or Windows environment variables to store API keys. This constitutes a persistence mechanism that alters the user's operating environment without a direct user command.- [DATA_EXPOSURE_AND_EXFILTRATION]: The Python script scripts/extract_ks_text.py scans multiple local shell configuration files (.zshrc, .bashrc, .bash_profile, .profile) to extract credentials. These credentials are then transmitted via network requests to the vendor's API endpoint at redfox.hk. While the domain is vendor-owned, the automated scanning of various shell profiles is a highly intrusive data access pattern.- [INDIRECT_PROMPT_INJECTION]: The skill processes transcription data (ASR) from external Kuaishou videos. This data is ingested into the agent's context, creating a surface for indirect prompt injection if the video audio contains instructions designed to manipulate the LLM. Ingestion Point: scripts/extract_ks_text.py (via API response). Boundary Markers: None. Capability Inventory: Network requests, file system read/write (shell profiles). Sanitization: No evidence of sanitization for the extracted ASR text.

kuaishou-video-extract