copilot-agent-patterns
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill includes 'Autonomy Guidance' in 'SKILL.md' that instructs agents to 'fix issues proactively' and specifically 'fix without asking' for security vulnerabilities. This creates a risk of autonomy abuse where the agent may perform unreviewed modifications to the repository, potentially hiding malicious changes under the guise of security fixes.
- [INDIRECT_PROMPT_INJECTION]: The design patterns in 'SKILL.md' describe agents that ingest untrusted data from repositories while possessing powerful capabilities.
- Ingestion points: Instructions to 'Analyze repository comprehensively' and 'Check existing codebase patterns' (SKILL.md).
- Boundary markers: The skill does not define specific delimiters to separate agent instructions from untrusted repository content (SKILL.md).
- Capability inventory: Patterns for agents include tools such as 'shell', 'edit', 'create', and 'search_code' (SKILL.md).
- Sanitization: There are no instructions for sanitizing or validating external repository content before processing (SKILL.md).
- [EXTERNAL_DOWNLOADS]: The skill references architectural guidelines and security policies from the vendor's official GitHub repository and established documentation sites.
Audit Metadata