The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes cluster diagnostic commands through a specialized wrapper, safe-kubectl.sh. This wrapper restricts actions to a specific allowlist of read-only verbs (e.g., get, describe, logs) and explicitly rejects mutating operations such as apply, delete, or exec. It also enforces security constraints on port-forwarding and proxy commands by binding them exclusively to the local loopback interface.
[EXTERNAL_DOWNLOADS]: The skill fetches supplementary documentation and indexes from official vendor-controlled domains, including arize.com and arize-ax.mintlify.site. These are treated as trusted sources for product-specific guidance.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data sources including Prometheus alert annotations and local documentation files. While this presents a potential surface for indirect prompt injection, the risk is mitigated by the skill's structured diagnostic workflow and the use of read-only access wrappers. Analysis of the ingestion points (alert APIs and local doc search) and sanitization (HTML tag stripping) indicates a low vulnerability profile.
[DYNAMIC_EXECUTION]: The verify-doc-links.py script dynamically loads logic from the sibling docs-search.py script using standard Python library mechanisms. This is used for consistent anchor parsing and does not involve executing untrusted or remote code.

arize-alerts-troubleshoot