The Agent Skills Directory

[PERSISTENCE]: The skill contains instructions for the agent to guide users in persisting the LINKFOX_AGENT_API_KEY by appending export commands to shell profile files such as ~/.zshrc and ~/.bashrc.
[OBFUSCATION]: The scripts/onboarding.py file contains a hardcoded Base64-encoded JSON string (eyJhX2lkIjoiNmEyMmM4YjA1YmM5MTZhIiwiZF9pZCI6IiJ9) used as a fixed identifier in login headers. Decoded, it contains non-malicious tracking IDs.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests untrusted user-supplied data (productTitle and productText) from product listings. 1. Ingestion points: Parameters productTitle and productText in SKILL.md and scripts/ruiguan_text_trademark_detection.py. 2. Boundary markers: Absent; the instructions do not specify delimiters to isolate user text. 3. Capability inventory: Execution of Python scripts, network requests to vendor APIs, and local file writes to the linkfox/ directory. 4. Sanitization: The Python script uses json.dumps() to safely serialize parameters before sending them to the API.
[EXTERNAL_DOWNLOADS]: The skill communicates with several vendor-owned domains (api.linkfox.com, agent-api.linkfox.com, and tool-gateway.linkfox.com) and utilizes third-party Python packages for its functionality.
[COMMAND_EXECUTION]: The skill relies on executing local Python scripts (scripts/ruiguan_text_trademark_detection.py and scripts/onboarding.py) to interact with its APIs and manage user accounts.

linkfox-ruiguan-text-trademark-detection