Opensearch Vector Search

Amazon OpenSearch vector search expert knowledge base. Comprehensive guidance on vector search configuration, cluster tuning, quantization, cost optimization...

This skill appears coherent with its stated purpose, but take these precautions before running anything or providing credentials: - Inspect the two scripts (scripts/analyze_cluster.py and scripts/get_opensearch_pricing.py) yourself or run them in an isolated environment to confirm they only perform read-only API calls and send data only to expected endpoints (your OpenSearch cluster and AWS Pricing API). - When asked to provide OpenSearch credentials, supply a least-privilege account (read-only, limited indices) rather than an admin user. Prefer short-lived credentials if possible. - For AWS pricing, use an IAM user/role with minimal permissions required for the Pricing API (or run the script locally under your own AWS profile) and rotate any credentials after use. - Do not allow the agent to automatically execute example curl/PUT/POST snippets from the references; treat those as documentation only. - Review network activity (outbound connections) when running the scripts to ensure no unexpected endpoints receive data. If you want additional assurance, paste the full contents of the two scripts here (or review them locally) and I can point out any lines that attempt to write, modify indices, or contact unexpected hosts. If you cannot review the code, consider running it on a disposable machine or container with no credentials first.

Type: OpenClaw Skill Name: opensearch-vector-search Version: 1.3.2 The skill bundle is a comprehensive expert system for Amazon OpenSearch vector search, providing documentation and diagnostic tools. It includes two Python scripts: `get_opensearch_pricing.py`, which uses the AWS Pricing API (boto3) to fetch public pricing data, and `analyze_cluster.py`, which performs read-only analysis of OpenSearch clusters. While the latter requires cluster credentials, the code and the instructions in `SKILL.md` are explicitly restricted to read-only operations (using GET and CAT APIs) for health checks and configuration reviews. No evidence of data exfiltration, malicious execution, or deceptive prompt injection was found; the functionality aligns entirely with the stated purpose of cluster optimization and capacity planning.