The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The installation instructions for the optional Rust CLI involve downloading a shell script from a remote URL and piping it directly into the bash interpreter (curl -fsSL https://raw.githubusercontent.com/volcengine/OpenViking/main/crates/ov_cli/install.sh | bash). This provides an avenue for arbitrary code execution from a source outside the primary vendor's infrastructure.
[REMOTE_CODE_EXECUTION]: The skill also suggests installing the CLI tool directly from an external Git repository (cargo install --git https://github.com/volcengine/OpenViking ov_cli), which involves downloading and executing code from a third-party source.
[EXTERNAL_DOWNLOADS]: The skill references and downloads components from the volcengine/OpenViking GitHub repository, which is an external third-party source.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external data (files, documentation) into the agent's workspace, creating a vulnerability surface for indirect prompt injection.
Ingestion points: External resources and documents are written to the workspace using the brain.write() method (e.g., resources/api_docs/stripe.md).
Boundary markers: The implementation examples do not show the use of delimiters or specific instructions to prevent the LLM from following commands embedded within the retrieved context.
Capability inventory: The skill uses retrieved content to construct prompts for LLM calls (call_llm), which are then executed by the agent.
Sanitization: There is no evidence of sanitization or validation of the ingested text before it is used to build the LLM's system prompt.

openviking-context-database