The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze runtime data from Power Automate flow executions, creating a surface where malicious data could influence agent behavior.
Ingestion points: External data is ingested through get_live_flow_run_action_outputs (specifically outputs.body) and get_live_flow_run_error (SKILL.md).
Boundary markers: The instructions do not define clear boundaries or provide "ignore embedded instructions" directives when processing flow outputs.
Capability inventory: The skill provides high-privilege capabilities including update_live_flow (to modify flow definitions) and trigger_live_flow (to initiate execution), which could be targets for an injection attack.
Sanitization: There is no evidence of content sanitization or validation for the data returned from flow actions before it is presented to the agent for analysis.

flowstudio-power-automate-debug