The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection because it ingests and processes untrusted data from external websites.
Ingestion points: Both scripts/extract.py and scripts/search.py fetch content directly from the public web (webpage content and search results).
Boundary markers: The skill output does not utilize clear delimiters or specific instructions to warn the agent that the fetched content should be treated as untrusted data.
Capability inventory: The agent has the ability to execute network operations via the included scripts and requests library.
Sanitization: The scripts perform deduplication and JSON formatting but do not filter or sanitize the actual text content for potentially malicious instructions hidden in the web data.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the requests library from the Python Package Index (PyPI) to facilitate API communication.
[COMMAND_EXECUTION]: The skill instructions involve executing local Python scripts (search.py and extract.py) to perform its primary functions. These scripts are invoked with user-supplied arguments (queries and URLs).

byted-byteplus-infoquest-search