improve-codebase-architecture

Pass

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests and processes untrusted codebase content.
  • Ingestion points: Untrusted data enters the context during the 'Explore' step where sub-agents navigate the codebase (SKILL.md).
  • Boundary markers: Absent. The instructions do not define delimiters or provide 'ignore' directives for content encountered in the codebase.
  • Capability inventory: The skill can spawn sub-agents and execute network-enabled commands like gh issue create (SKILL.md).
  • Sanitization: Absent. There is no evidence of escaping or filtering codebase content before it is processed or included in GitHub issues.
  • [COMMAND_EXECUTION]: The skill uses the gh CLI tool to create issues on GitHub. The instructions in SKILL.md (Step 7) explicitly state: 'Do NOT ask the user to review before creating — just create it and share the URL.' This practice of skipping user confirmation for automated external actions reduces oversight and increases the risk of unintended or malicious data being posted to public or shared repositories.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 7, 2026, 02:53 PM
Security Audit — agent-trust-hub — improve-codebase-architecture