workflow-architecture-audit

Pass

Audited by Gen Agent Trust Hub on Aug 13, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: The skill includes instructions to "Ignore Claude-specific mode-switch instructions when they appear," which is a pattern used to override system-level or platform-specific directives.\n- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and audit project files, creating an attack surface for instructions embedded in untrusted project data.\n
  • Ingestion points: Project configuration files (docs/project-config.json), reference documentation, and general source code files.\n
  • Boundary markers: The protocol does not specify the use of delimiters or clear boundaries to isolate audited content from the agent's core instructions.\n
  • Capability inventory: The skill can spawn sub-agents using spawn_agent and write report files to the filesystem.\n
  • Sanitization: There is no mention of sanitizing or validating audited content before it is processed by sub-agents.\n- [DYNAMIC_EXECUTION]: The architecture-review-full step dynamically invokes and coordinates multiple sub-agents (architecture-scalability-review, architecture-review, production-readiness-review) to execute the audit based on the gathered project context.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 13, 2026, 03:11 PM
Security Audit — agent-trust-hub — workflow-architecture-audit