The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill ingests data from external sources and local project files to generate code and metadata, which could potentially be used to influence agent behavior if the source content is compromised.
Ingestion points: Flow names, descriptions, and IDs from the Power Automate API, as well as existing project frontend code and web role definitions.
Boundary markers: The skill does not use specific delimiters or explicit instructions to the agent to ignore embedded commands within the ingested data.
Capability inventory: The skill employs Write, Edit, and Bash tools to modify the file system and generate executable scripts.
Sanitization: Basic HTML escaping is implemented in the plan rendering template, and the metadata generation script uses single-quoting for YAML scalar values to prevent structural corruption.
Network Operations: The skill fetches flow and environment data from Microsoft's Power Automate service endpoints (e.g., api.flow.microsoft.com). This is standard functionality for listing and identifying flows to be integrated.
Command Execution: The skill executes local helper scripts and CLI tools (such as pac auth who) to manage authentication, list environment resources, and generate metadata files.

add-cloud-flow