improve-codebase-architecture

Pass

Audited by Gen Agent Trust Hub on Apr 2, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted data from the codebase.
  • Ingestion points: Reads files and navigates the codebase using the Explore subagent tool as described in SKILL.md.
  • Boundary markers: The skill lacks explicit instructions or delimiters to isolate content extracted from the codebase when passing it to sub-agents or generating RFCs.
  • Capability inventory: The skill uses gh issue create and createJiraIssue to write data externally, and spawns multiple sub-agents for parallel processing.
  • Sanitization: No validation or sanitization of the codebase content is performed before it is incorporated into the generated issue descriptions.
  • [PROMPT_INJECTION]: The skill instructs the agent to bypass human oversight for external write operations.
  • Evidence: SKILL.md explicitly states, 'Do NOT ask the user to review before creating — just create it and share the URL.' This instruction encourages autonomous execution which can lead to the accidental posting of sensitive information or malicious content derived from a poisoned codebase.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 2, 2026, 10:34 PM
Security Audit — agent-trust-hub — improve-codebase-architecture