The Agent Skills Directory

[External Data Ingestion]: The skill fetches content from Microsoft Learn (a well-known service) to ground its deployment logic in current documentation. While this brings external data into the agent's context, the skill is designed to surface any detected divergences to the user as a warning before proceeding with deployment actions.
[Environment Security Configuration]: The skill can modify environment-level 'blockedattachments' settings to ensure Power Pages bundle files (like .js) can be imported. Because this change has tenant-wide implications, it is strictly managed through explicit consent gates (AskUserQuestion) during both pre-flight and reactive failure handling phases.
[Indirect Prompt Injection Surface]: The skill ingests untrusted data from external documentation and AI-generated deployment notes retrieved from the Dataverse API. This creates a surface for indirect instructions, although the impact is mitigated by the skill's focus on structured deployment parameters and the requirement for user approval at critical stages.
[Command Line Operations]: The skill utilizes the Bash tool to execute standard CLI utilities (az CLI, PAC CLI, and git) alongside local helper scripts. These operations are necessary for the skill's Application Lifecycle Management (ALM) purpose and rely on the user's existing authenticated sessions and environment permissions.

deploy-pipeline