copilot-cli

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructions and reference materials describe the execution of the copilot and gh (GitHub) CLI tools via the Bash tool to delegate coding tasks to external models.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied requests and project context, which are passed to the GitHub Copilot CLI for processing. This creates a surface for indirect prompt injection if the external model output contains instructions intended for the agent.
  • Ingestion points: User input and task descriptions are ingested in SKILL.md (Step 2) to be converted into English prompts.
  • Boundary markers: The skill provides a prompt template including "Constraints" and "Expected output" sections to help delimit the task scope.
  • Capability inventory: The skill utilizes the Bash tool for CLI command execution and Read/Write tools for file operations and output sharing as described in SKILL.md.
  • Sanitization: The instructions include a "Constraints and Warnings" section explicitly cautioning against including secrets in prompts and advising users to validate external output before applying changes.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 09:08 PM
Security Audit — agent-trust-hub — copilot-cli