splunk-platform（斯普伦克平台）

Splunk Platform

Use this as the default skill for Splunk work. It should answer most Splunk framework-selection questions directly and send you to exactly one or two reference files for implementation details.

Read only the references that match the task:

• references/python-sdk.md for Python automation, splunklib, and result parsing
• references/javascript-sdk.md for Node/browser JS SDK work
• references/rest-search-patterns.md for raw REST, search jobs, and export patterns
• references/admin-searches.md for read-only admin/discovery SPL
• references/ucc-framework.md for add-ons, modular inputs, setup pages, and alert actions
• references/dashboard-development.md for Dashboard Studio and Simple XML
• references/mcp-integration.md for agent-facing Splunk tool design
• references/platform-admin.md for install/upgrade/deployment automation
• references/app-packaging.md for AppInspect, packaging, and release hygiene

Start Here

Classify the task before you write code:

1. External automation: Python or JS code talks to Splunk over REST/SDK.
2. Search/discovery: SPL inspects indexes, metadata, users, apps, and knowledge objects.
3. App/add-on engineering: packaged Splunk app, technical add-on, modular input, alert action, setup UI.
4. Dashboards/UI: Dashboard Studio JSON, Simple XML, or legacy SplunkJS/Web Framework.
5. AI integration: MCP server or other agent-facing tools over Splunk.
6. Platform administration: host deployment, upgrades, distributed topology, app rollout.

If the task spans multiple areas, pick the primary deliverable first. A script that queries Splunk is external automation, not an add-on.

If the user asks whether a Splunk analysis plan "maps to best practices", asks for "Tier 0" or "Tier 1", or wants to normalize an analytics repo around Splunk-backed exploration, answer from the perspective of a full analysis workflow, not just Splunk infrastructure.

Strong Defaults

• Prefer Python for automation, exports, CLIs, notebooks, and agent backends.
• Prefer JavaScript SDK only when the surrounding system is already Node/JS or you are in Splunk web-facing code.
• Prefer raw REST when you need streaming export semantics, exact endpoint control, or unsupported SDK behavior.
• Prefer UCC for new technical add-ons. Do not hand-roll setup pages and REST handlers unless you are maintaining an existing non-UCC app.
• Prefer Dashboard Studio for new dashboards.
• Prefer Simple XML only for legacy maintenance or when existing app behavior is tightly coupled to XML/tokens.
• Prefer read-only SPL for discovery and audits.
• Prefer narrow MCP tools over one generic "run any SPL" endpoint.
• Prefer official automation repos for platform deployment before inventing custom shell glue.
• Run AppInspect/package validation before claiming an app or add-on is shippable.

What Not To Use

• Do not build a Splunk app when the real need is an external export script.
• Do not use browser-side JS SDK code to hold long-lived credentials unless there is no server-side alternative.
• Do not default to oneshot for large searches.
• Do not expose unrestricted search execution to LLMs.
• Do not create new HTML dashboards or lean on deprecated web framework patterns for greenfield work.
• Do not use write-side SPL commands in automation unless the user explicitly wants state changes.
• Do not assume Splunk Cloud lets you use every REST/admin path that Splunk Enterprise does.

Decision Table

Need data out of Splunk for analysis, ETL, or a CLI?

Use python-sdk.md and rest-search-patterns.md.

Default: Python SDK for auth/job lifecycle, raw REST export for large streaming result sets.

Need to inspect a Splunk instance, inventory objects, or audit config?

Use admin-searches.md.

Default: rest, metadata, and tstats. Avoid raw event scans unless you need event content.

Need a Splunk add-on with config UI, modular inputs, or alert actions?

Use ucc-framework.md and likely app-packaging.md.

Default: UCC. Treat packaging/AppInspect as part of the implementation, not postscript.

Need a dashboard or dashboard migration?

Use dashboard-development.md.

Default: Dashboard Studio for new work; Simple XML for edits inside an existing XML-heavy app.

Need an MCP server or AI-safe integration?

Use mcp-integration.md plus either python-sdk.md or rest-search-patterns.md.

Default: server-side credentials, read-only-by-default tools, validated SPL.

Need to install, upgrade, or automate Splunk infrastructure?

Use platform-admin.md.

Default: official Splunk automation repos and admin manual concepts, not bespoke scripts first.

Search Execution Defaults

• Add explicit time bounds.
• Add explicit limits or paging.
• Use search/jobs for managed jobs.
• Use export endpoints when you need streaming output and do not need a persistent SID.
• Use SDK job abstractions for moderate searches where polling and result paging are acceptable.
• Keep query construction separate from result handling.

Enterprise Vs Cloud

• Splunk Enterprise gives broader host/admin access.
• Splunk Cloud often constrains platform-level operations and may require Support enablement for specific REST/API capabilities.
• For app/add-on guidance, always check whether the task is Cloud-safe before recommending local filesystem or admin-server assumptions.

Knowledge Object Defaults

Treat these as first-class assets:

• saved searches and alerts
• dashboards/views
• macros
• lookups
• field extractions and props/transforms-driven behavior
• data models
• KV store collections

For inventory and audits, start with admin-searches.md. For packaging and app delivery, include those objects intentionally in the app structure and validate them with app-packaging.md.

Safe Patterns

• Keep auth in env vars or approved secret stores.
• Separate read paths from write paths in code and tool design.
• Scope namespaces deliberately when using REST or SDK config/object APIs.
• Return structured output from automation and MCP tools.
• When in doubt, choose the path that is easiest to reason about operationally: Python script > custom REST endpoint > full Splunk app.

Replaces

This skill subsumes the old:

• splunk-sdk
• splunk-quick-searches

Keep the shims for backward compatibility, but maintain real guidance here.