orchestrate-batch-refactor

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill instructions define a robust and standard architectural pattern for managing complex code changes. It emphasizes safety through plan synthesis, dependency awareness, and integration testing.\n- [INDIRECT_PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection as it requires agents to ingest and analyze potentially untrusted code from a repository.\n
  • Ingestion points: Sub-agents (explorers) are instructed to analyze the target codebase scope (paths and modules) which could contain malicious comments or code patterns designed to mislead the agents.\n
  • Boundary markers: The skill explicitly instructs sub-agents to ignore unrelated edits and stay within their owned files as defined in the worker prompt template.\n
  • Capability inventory: The workflow involves sub-agents modifying repository files and executing validation commands such as tests or type-checks.\n
  • Sanitization: Safety is managed through a multi-stage integration process and manual review gates rather than automated sanitization of the code being refactored.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 12:44 PM
Security Audit — agent-trust-hub — orchestrate-batch-refactor