copilot-studio-adaptive-card-builder
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Command Execution: The skill instructions direct the agent to execute a local Python script (
scripts/validate_cards.py) to perform semantic linting on Adaptive Card artifacts. This is an intended feature designed to ensure the generated JSON meets platform requirements and safety policies before being provided to the user. - Security Auditing Logic: The bundled linter script contains defensive logic to detect and prevent security risks in UI artifacts, such as fields that attempt to collect credentials (passwords, API tokens) or perform destructive operations without explicit confirmation toggles.
- Indirect Prompt Injection Surface: The skill is designed to ingest and process Adaptive Card JSON provided by users. It implements significant mitigations against potential parsing or injection attacks, including iterative depth-limit checks for JSON objects to prevent resource exhaustion and strict schema validation.
- Defensive Design: The linter script follows security best practices by using the Python standard library exclusively, implementing a custom JSON depth gate before parsing, and employing tokenization-based matching to identify sensitive identifiers in card metadata.
Audit Metadata