find-bugs

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data in the form of code diffs and file contents, which presents an attack surface for indirect prompt injection.
  • Ingestion points: The skill ingests data from git diff master...HEAD and reads the full content of modified files in Phase 1.
  • Boundary markers: The instructions do not define clear boundary markers or provide the agent with guidance to ignore instructions embedded within comments, string literals, or documentation in the files being reviewed.
  • Capability inventory: The skill uses read-only tools to examine local files and git history. It does not explicitly request network access or write permissions, though it generates a report for the user.
  • Sanitization: There is no explicit sanitization or filtering of the ingested content before it is processed by the LLM for analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 02:31 PM
Security Audit — agent-trust-hub — find-bugs