The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references and downloads official configuration manifests and deployment charts from the Milvus project's GitHub repository and Zilliz's Helm repository. These are well-known, trusted sources for the software. Detailed URLs include: https://github.com/milvus-io/milvus/releases/download/v3.0-beta/milvus-standalone-docker-compose.yml and https://zilliztech.github.io/milvus-helm/.
[COMMAND_EXECUTION]: The documentation includes standard commands for installation and deployment, such as pip install, wget, docker compose, and helm. A provided utility script, scripts/check-updates.py, verifies the skill's integrity and checks for upstream version updates by querying PyPI (https://pypi.org/pypi/pymilvus/json). This script performs read-only operations on local skill files and does not execute arbitrary remote code.
[INDIRECT_PROMPT_INJECTION]: The skill documents Retrieval-Augmented Generation (RAG) patterns that ingest external data into Milvus collections for later retrieval by an LLM.
Ingestion points: Untrusted external data enters the system through insertion commands (e.g., client.insert, Milvus.from_documents) documented in references/03-data-operations.md and references/11-integrations.md.
Boundary markers: The RAG prompt examples use simple string interpolation (e.g., "Answer using context:\n{context}"). While common in tutorials, these lack robust delimiters or specific safety instructions to isolate retrieved content.
Capability inventory: The skill utilizes file system access for local SQLite-style databases and network operations to communicate with Milvus server endpoints.
Sanitization: The provided examples do not demonstrate explicit sanitization of external text before vectorization or inclusion in prompts.

milvus