m365-agent-developer

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • COMMAND_EXECUTION (HIGH): Indirect Prompt Injection Surface. The skill provides the agent with high-privilege capabilities to execute local commands (npm install, npm run compile, atk provision). The 'Critical Workflow Rules' in SKILL.md require the agent to automatically redeploy after any code changes derived from user requirements. This lack of a manual review step between processing untrusted user input and executing deployment commands on the host system represents a significant security risk.\n
  • Ingestion points: User requirements gathering in Step 1 and TypeSpec code implementation in Step 3 of SKILL.md.\n
  • Boundary markers: No specific boundary markers or instruction-ignoring delimiters are defined for user-provided requirements or code blocks.\n
  • Capability inventory: Execution of npm run compile and atk provision via npx (documented in SKILL.md, references/examples.md, and README.md).\n
  • Sanitization: No sanitization or validation of user-provided TypeSpec code or requirements is performed before execution.\n- EXTERNAL_DOWNLOADS (LOW): The skill instructs the agent to download and execute code from a remote source via npx.\n
  • Evidence: Use of npx -p @microsoft/m365agentstoolkit-cli@latest atk provision in SKILL.md and README.md.\n
  • Trusted source: The package resides within the @microsoft scope, which is a verified trusted organization. The finding is downgraded to LOW per the [TRUST-SCOPE-RULE].
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 02:15 AM