Add purview-developer skill

[yulcodeoy]

Summary

Adds a new purview-developer skill that teaches AI coding agents how to correctly build on the Microsoft Purview Developer Platform.

Scenarios covered

1. Reading and applying sensitivity labels — listing tenant labels via GET /security/dataSecurityAndGovernance/sensitivityLabels
2. Logging contentActivity events — creating audit records via POST /users/{id}/dataSecurityAndGovernance/activities/contentActivities
3. Reading governance state — computing protection scopes and processing content via protectionScopes/compute and processContent APIs
4. Agent Framework integration — wiring Purview middleware into Agent Framework agents using Microsoft.Agents.AI.Purview (C#) and agent-framework-purview (Python)

Key features

• Decision guidance table: answers "which Purview surface do I call?" — Graph APIs directly vs. Agent Framework middleware
• Working C# and Python examples for all four canonical scenarios
• Anti-patterns section: common mistakes agents generate (wrong endpoints, legacy APIs, double-counting, wrong auth scopes, missing caching, ignoring execution mode) with correct replacements
• "When NOT to use" section: clarifies this is for developers, not admin/portal tasks

References

All content grounded in these authoritative sources:

• https://learn.microsoft.com/en-us/purview/developer/
• https://github.com/microsoft/purview-api-samples
• https://learn.microsoft.com/en-us/graph/security-datasecurityandgovernance-overview
• https://learn.microsoft.com/en-us/graph/api/resources/contentactivity?view=graph-rest-1.0
• https://learn.microsoft.com/en-us/graph/api/resources/security-sensitivitylabel?view=graph-rest-1.0
• https://learn.microsoft.com/en-us/graph/api/resources/tenantdatasecurityandgovernance?view=graph-rest-1.0
• https://learn.microsoft.com/en-us/graph/api/resources/userdatasecurityandgovernance?view=graph-rest-1.0
• https://learn.microsoft.com/en-us/agent-framework/integrations/purview?pivots=programming-language-csharp
• https://github.com/microsoft/agent-framework/tree/main/dotnet/src/Microsoft.Agents.AI.Purview
• https://github.com/microsoft/agent-framework/tree/main/python/packages/purview

[thegovind]

Review summary

In-scope, well-structured Microsoft Purview developer skill. Authoritative coverage of Graph APIs, Agent Framework Purview middleware, and Agent 365 telemetry. A few small things to address before merge.

Issues

🟡 Recommended

• README catalog not updated. This PR adds a new core skill but the diff only contains the two skill files. README.md lists skill counts ("Core: 6", total of 130) — please bump those numbers and add the entry to the Core section so the skill is discoverable in the catalog.
• Symlink not created. Other Microsoft-platform core skills appear under skills/<lang>/<category>/ via symlink (per AGENTS.md and the established pattern). Since this skill ships C# and Python samples, consider symlinking from skills/dotnet/security/purview and skills/python/security/purview (or similar category) so it surfaces in language-scoped browsing.
• Tests/scenarios not included. The repo convention (see tests/scenarios/<skill>/scenarios.yaml for credential-free-dev, entra-agent-id, etc.) is to ship scenario tests alongside new skills. For a security/compliance skill, scenario coverage matters more than usual — please add at minimum a happy-path scenario per surface (read labels, log contentActivity, processContent, middleware wiring).
• Description length. The description field is ~990 characters with a long inline trigger list. That works, but it's loaded into context every time the agent decides whether to invoke the skill. Consider trimming to one paragraph + 8-12 highest-signal triggers, and moving the long trigger inventory into a references/triggers.md if needed.

🟢 Nits

• L74 / Scenario 1 GET example: the Authorization: Bearer {token} placeholder is fine for HTTP exposition, but please add a one-liner reminder that production callers should use DefaultAzureCredential + GraphServiceClient (which the C#/Python sub-sections then do correctly) so a copy-paste reader doesn't end up token-juggling by hand.
• Line 561 / Scenario 4 (InteractiveBrowserCredential): worth adding a one-line note that InteractiveBrowserCredential is a developer/desktop scenario; production agent hosts should use managed identity via DefaultAzureCredential (or ManagedIdentityCredential).
• Mention of "Microsoft 365 E5 license (or equivalent) with Microsoft Purview configured and pay-as-you-go billing" under Prerequisites is helpful — please double-check the licensing wording against current Microsoft Learn so the skill stays accurate when it's loaded into agent context.

Solid contribution overall — once the catalog entry, symlinks, and scenarios are added, this is ready.

[yulcodeoy]

@microsoft-github-policy-service agree company="Microsoft"