The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions to install the zilliz-cli by fetching and executing a script directly from the vendor's official GitHub repository. As this is a vendor-owned resource for its own management skill, it is documented as a functional requirement.
Evidence: curl -fsSL https://raw.githubusercontent.com/zilliztech/zilliz-cli/master/install.sh | bash in SKILL.md and references/setup.md.
[COMMAND_EXECUTION]: The skill's core functionality involves executing shell commands via the zilliz CLI to manage vector database clusters, collections, and data operations.
[INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection because it processes user-provided data and search filters which are then used as arguments in CLI commands.
Ingestion points: The --data flag for vector insertions and searches, the --filter flag for scalar queries, and the --body flag for complex JSON configurations throughout the reference files.
Boundary markers: The skill instructs the agent to use specific command-line flags and expects the agent to use --output json for programmatic parsing, but it does not specify explicit prompt delimiters (like xml tags) for user data.
Capability inventory: The skill allows for various cluster and data management operations, including destructive ones like drop and delete (which require explicit user confirmation per the safety rules).
Sanitization: No explicit prompt-level sanitization (e.g., escaping or validation) of the user-provided strings is described; the CLI tool itself is expected to handle input validation at the execution layer.

zilliz