The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the openviking Python package via pip to provide its core functionality.
[COMMAND_EXECUTION]: The viking.sh wrapper script executes dynamic Python code using the python3 -c command. It employs environment variables to pass arguments safely, which mitigates direct shell injection into the Python environment.
[DATA_EXFILTRATION]: The skill reads local files and directories provided by the user to build a semantic search index. This data is sent to external embedding providers (such as NVIDIA) as part of its normal operation.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection because it ingests untrusted local data that may be retrieved and processed by the AI agent.
Ingestion points: File and directory paths processed by the add and add-dir commands in viking.sh.
Boundary markers: No explicit delimiters or instruction-ignore markers are used when processing file content.
Capability inventory: File system read access, semantic search, and document summarization.
Sanitization: No explicit sanitization or filtering of file content is performed within the provided scripts.

openviking