Lesson 9 — Understanding Security Copilot Plugins
Plugins connect Microsoft Security Copilot to security products, intelligence services, websites, APIs and organisation-specific workflows.
They provide the skills and data sources Security Copilot can use when responding to a prompt, but they remain subject to configuration, authentication, product permissions and governance.
This lesson explains Microsoft, non-Microsoft, website and custom plugins, how Security Copilot selects plugin skills and how analysts validate every plugin-assisted result.
What you will learn
This lesson explains how plugins extend Security Copilot and how analysts manage, govern and validate them.
Security Copilot plugin model
↓
Security Copilot interprets the requested task
↓
Enabled plugins and available skills are evaluated
↓
A relevant plugin skill is selected
↓
Authentication and permissions are checked
↓
The plugin retrieves data or performs its configured function
↓
Plugin output is grounded into the response
↓
The process log records the selected source and action
↓
The analyst opens the original data and validates the result
Plugin categories
| Category | Purpose | Primary consideration |
|---|---|---|
| Microsoft | Supported Microsoft security and IT services. | Usually uses on-behalf-of authentication and existing product permissions. |
| Non-Microsoft | Approved third-party security or intelligence services. | May require separate authentication and third-party terms. |
| Websites | Supported website-based information sources. | Review trust, relevance and data-processing implications. |
| Custom | Organisation-created skills, APIs, KQL or specialised workflows. | Requires manifest design, testing, ownership and governance. |
Learning objectives
- Define Security Copilot plugins and skills.
- Recognise the four plugin categories.
- Manage plugins through Sources.
- Understand authentication and permissions.
- Recognise custom plugin structure.
- Reduce plugin ambiguity and risk.
- Validate plugin-assisted output.
What is a plugin?
A plugin extends Security Copilot by exposing a supported data source, function or security capability.
Plugins provide skills
Each plugin contains one or more skills that describe the operations Security Copilot can invoke.
Plugins are not passive labels
An enabled plugin can influence which capability is selected and what external context becomes available.
Microsoft plugins
Microsoft plugins connect Security Copilot to supported Microsoft security products and services.
Non-Microsoft plugins
Non-Microsoft plugins connect approved third-party security services and intelligence sources.
Website plugins
Website plugins can provide controlled access to supported web-based information sources.
Custom plugins
Custom plugins allow organisations to extend Security Copilot for specialised tasks and internal workflows.
Manage plugins through Sources
The Sources control in the prompt bar opens plugin management and file-source settings.
Enable or disable plugins
Turn plugins on when their capabilities are needed and off when tighter scope is useful.
Use filters
Filter plugins by state and category to find active, inactive, Microsoft, non-Microsoft or custom sources.
Restricted plugins
Preinstalled plugins the user cannot access can appear as Restricted.
Custom plugins may be hidden
Custom plugins the user cannot access do not appear in the list.
Personalise plugin settings
Some plugins support user-specific settings such as a default Microsoft Sentinel workspace.
Workspace selection matters
A Sentinel prompt can return the wrong context when the default workspace is incorrect.
On-behalf-of authentication
Active Microsoft plugins generally access security data on behalf of the signed-in user.
Product permissions remain authoritative
Security Copilot cannot use a Microsoft plugin to retrieve data the user is not authorised to access.
Plugin-specific authentication
Some plugins require their own API key, token or authentication setup instead of on-behalf-of access.
Authentication scope matters
A plugin credential should grant only the permissions required for the approved task.
Plugin output can be incomplete
Missing permissions, configuration or source data can produce partial responses.
Plugin selection is prompt-driven
Security Copilot uses the prompt and plugin descriptions to select an appropriate skill.
Descriptions guide selection
Clear skill descriptions help the model distinguish when one capability should be used instead of another.
Skill collision
A skill collision occurs when Security Copilot cannot reliably distinguish between similar plugin skills.
Reduce skill collision
Use distinct descriptions and consolidate overlapping skills where practical.
Supporting plugins
Some prompts require more than one plugin or supporting capability to produce a complete response.
Check the process log
The process log reveals which plugin, skill, source and action were selected.
Plugin choice is not proof
The selected plugin may be correct while the resulting interpretation is still wrong.
Microsoft Defender plugins
Defender plugins can expose incidents, alerts, entities, devices, mail and other supported security context.
Microsoft Sentinel plugin
The Sentinel plugin supports workspace-based investigation, incident and KQL scenarios.
Microsoft Entra plugin
Entra capabilities can support identity, sign-in and access investigations.
Microsoft Intune plugin
Intune capabilities can support device, user, compliance and configuration questions.
Microsoft Purview plugin
Purview capabilities can support data-security and compliance workflows.
Threat intelligence plugins
Threat-intelligence plugins enrich indicators, actors, campaigns and infrastructure.
Third-party plugin risk
Third-party plugins introduce separate data-processing, privacy, availability and contractual considerations.
Review third-party terms
Organisations remain responsible for configuring and using third-party integrations lawfully and appropriately.
API keys are sensitive
Store and rotate plugin secrets according to organisational credential-management standards.
Data leaves the source system
Understand what data is retrieved, where it is processed and how long it is retained.
Custom plugin manifests
Custom plugins use YAML or JSON manifests describing metadata and how skills are invoked.
Descriptor metadata
The descriptor identifies the plugin and provides human-readable and model-facing descriptions.
Skill groups
Skill groups define the functions or tools the plugin makes available.
API plugins
API-based plugins can call supported external APIs through a defined specification.
KQL plugins
KQL-based custom capabilities can expose approved parameterised queries for repeatable analysis.
GPT plugins
GPT-based plugin skills can apply structured prompt templates to specialised tasks.
MCP plugins
Model Context Protocol plugins can connect compatible tools and context through supported Security Copilot integration.
Custom plugin governance
Owners should control whether contributors can add custom plugins to their sessions.
Test before publishing
Validate custom plugins with representative, negative and failure test cases.
Validate parameters
Ensure entity names, time ranges, IDs and other inputs are passed correctly.
Handle errors safely
Plugin errors should remain visible and should not be converted into confident conclusions.
Version plugin manifests
Track changes to endpoints, descriptions, authentication, fields and expected outputs.
Retire outdated plugins
Disable integrations that rely on deprecated APIs, old schemas or unsupported workflows.
Limit plugin count
Enabling every available plugin can increase ambiguity and make source selection harder to understand.
Use task-specific sources
Select only the plugins relevant to the current investigation when practical.
Start a clean session after changes
Begin a new session after materially changing plugin configuration to reduce stale context.
Document dependencies
Promptbooks and agents should identify the plugins and permissions they require.
Plugins and agents work together
Agents can use plugins as tools to retrieve context or perform configured tasks.
Plugins and promptbooks work together
Promptbook prompts can depend on plugin skills and data returned earlier in the sequence.
Review source attribution
Final reports should identify which product or plugin supplied important evidence.
Human validation remains essential
Plugins expand access and capability, but analysts still validate every security conclusion.
Plugin selection workflow
2. Identify the required security data
3. Open Sources
4. Enable the relevant plugin
5. Confirm authentication and permissions
6. Configure required plugin settings
7. Start a clean session
8. Submit a focused prompt
9. Review the selected skill in the process log
10. Open the original product or intelligence source
11. Validate returned entities, timestamps and context
12. Record plugin limitations and confidence
Agent Foskett investigation: “The right plugin used the wrong workspace…”
↓
The Microsoft Sentinel plugin was enabled
↓
The process log showed the expected plugin skill
↓
The response stated that no matching incidents existed
↓
The analyst knew an incident had been created that morning
↓
Agent Foskett opened the plugin settings
↓
The default workspace pointed to the development environment
↓
The production workspace contained the incident
↓
The setting was corrected
↓
A clean session was started
↓
The prompt returned the expected incident
↓
The plugin had worked exactly as configured
↓
The configuration had supplied the wrong context
↓
Plugin validation solved the mystery before another query was written
Plugin validation checklist
| Area | Question | Validation action |
|---|---|---|
| Category | Is this Microsoft, non-Microsoft, website or custom? | Apply the correct governance controls. |
| Status | Is the plugin enabled and available? | Check Sources and restrictions. |
| Authentication | How does the plugin authenticate? | Confirm user permissions or plugin credentials. |
| Settings | Is the correct workspace, tenant or scope configured? | Review personalised settings. |
| Skill | Which skill answered the prompt? | Inspect the process log. |
| Input | Were IDs, entities and time ranges passed correctly? | Validate parameters. |
| Source | Does the original source contain the returned evidence? | Open and compare the record. |
| Third party | What external terms and data processing apply? | Review contractual and privacy obligations. |
| Error | Did a failure remove required context? | Do not infer success from missing output. |
| Confidence | Does the evidence support the conclusion? | Document limitations and human validation. |
Key takeaways
- Plugins extend Security Copilot with data sources, functions and specialised security capabilities.
- Plugin categories include Microsoft, non-Microsoft, websites and custom.
- Plugins can be enabled, disabled, filtered and personalised through Sources.
- Microsoft plugins generally use on-behalf-of authentication and preserve product permissions.
- Some non-Microsoft and custom plugins require separate authentication.
- Restricted or hidden plugins reflect access and visibility controls.
- Custom plugins use YAML or JSON manifests that describe metadata and skills.
- Clear descriptions help Security Copilot select the correct skill and reduce collisions.
- Third-party plugins introduce separate privacy, contractual and operational risks.
- Every plugin-assisted result must be validated against the original source.
What Agent Foskett checked
- Plugin category
- Enabled status
- User permissions
- Authentication method
- Default workspace
- Selected skill
- Process log
- Original incident
- Session context
- Final confidence
Best practices
- Enable relevant plugins only.
- Apply least privilege.
- Protect API keys.
- Review plugin settings.
- Use clear skill descriptions.
- Test custom plugins.
- Version manifests.
- Review third-party terms.
- Inspect process logs.
- Validate original evidence.
Related Agent Foskett resources
Continue the Microsoft Security Copilot Academy
What are Microsoft Security Copilot plugins?
Security Copilot plugins connect the platform to Microsoft, non-Microsoft, website and custom security capabilities and data sources.
How do Security Copilot plugins work?
Enabled plugin skills are selected according to the prompt, authentication and permissions, with actions and sources visible in the process log for analyst validation.
Custom Security Copilot plugins
Custom plugins use YAML or JSON manifests to define metadata, skill groups, authentication and invocation details for specialised organisational workflows.
