oh-interfaces-ipc-to-service

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHPROMPT_INJECTION
Full Analysis
  • [Indirect Prompt Injection] (HIGH): The skill creates an attack surface where malicious user input can influence the generation of security-sensitive system code.
  • Ingestion points: User-provided C++ interface declarations and authentication level specifications (Steps 1 and 2 in SKILL.md).
  • Boundary markers: Absent. The instructions do not include delimiters or warnings for the agent to ignore potentially malicious logic embedded within the user's interface declaration.
  • Capability inventory: The agent is instructed to modify multiple files across a Client/Service architecture, including the IPC Proxy, Stub, and most critically, the IsExclusiveVerificat function in the access verifier script which handles IsSystemCalling or IsFoundationCalling (Steps 6 and 7).
  • Sanitization: Absent. While the skill asks the agent to check for C++ 17 syntax and style consistency, there are no instructions to validate or sanitize the logic of the user-provided code snippets before incorporating them into the system stubs and verifiers.
  • [Metadata Poisoning] (LOW): The description accurately reflects the skill's purpose, but the instructional nature of the markdown body relies entirely on the agent's ability to safely interpret user-provided data for system-level modifications.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 12:43 PM