copilot-sdk
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [Indirect Prompt Injection Surface]: The skill facilitates the development of agentic workflows that process user-provided prompts and external tool outputs. This architecture presents a potential surface for indirect prompt injection.
- Ingestion points: User inputs are ingested via CLI prompts in the example code snippets.
- Boundary markers: The current documentation examples do not demonstrate the use of delimiters or specific instructions to isolate external data from system instructions.
- Capability inventory: The SDK enables tool definition and connection to MCP servers, providing a functional surface for the agent to perform actions.
- Sanitization: No explicit input sanitization or data validation logic is shown in the foundational examples.
- [External Dependency References]: The skill documentation includes instructions to install packages from public registries. These are official development kits provided by the vendor.
- Evidence: Installation commands for
@github/copilot-sdkandgithub-copilot-sdkfound inSKILL.md.
Audit Metadata