improve-codebase-architecture

Pass

Audited by Gen Agent Trust Hub on May 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the GitHub CLI tool (gh issue create) to automate the creation of RFCs as issues on remote repositories.
  • [DATA_EXFILTRATION]: The skill instructions explicitly bypass human oversight before external communication ('Do NOT ask the user to review before creating
  • just create it and share the URL'). This creates a risk of exposing sensitive implementation details or internal documentation to potentially public GitHub issue trackers if the agent includes such information in the generated RFC body.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. 1. Ingestion points: The agent is instructed to explore the local codebase files (SKILL.md, Step 1). 2. Boundary markers: The instructions lack markers or warnings to disregard malicious instructions embedded in the codebase being analyzed. 3. Capability inventory: The skill can create GitHub issues and spawn multiple autonomous sub-agents with specific design briefs. 4. Sanitization: There is no evidence of filtering or sanitizing codebase content before it is processed by sub-agents or included in the final GitHub issue payload.
Audit Metadata
Risk Level
SAFE
Analyzed
May 12, 2026, 01:34 PM
Security Audit — agent-trust-hub — improve-codebase-architecture