gathering-architecture

Pass

Audited by Gen Agent Trust Hub on Oct 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied system specifications to drive a multi-stage architectural design and code generation workflow.
  • Ingestion points: The 'SUMMON' phase (Phase 1) in SKILL.md ingests user-provided problem statements, scale requirements, and constraints.
  • Boundary markers: The workflow uses strict context isolation; subagents (Eagle, Crow, Swan, Elephant) receive specific, filtered inputs to ensure fresh-eyes evaluation. For example, the 'Crow' receives only the architecture document, and the 'Elephant' receives only the technical specification.
  • Capability inventory: The 'Elephant' subagent in SKILL.md (Phase 5) is designed to create and modify architectural foundation files (database migrations, services, API routes, and configuration files) and executes verification commands (pnpm install, gw dev ci).
  • Sanitization: The instructions in references/conductor-dispatch.md mandate the use of 'Rootwork' type safety utilities and Zod schema validation for all data boundaries, JSON/KV reads, and form inputs.
  • [DYNAMIC_EXECUTION]: The skill utilizes a multi-agent orchestration pattern to dynamically generate technical artifacts and code foundations based on user specifications. This is a core feature of the architecture gathering workflow.
  • [COMMAND_EXECUTION]: The verification gates in SKILL.md execute shell commands (pnpm install, gw dev ci) to provision dependencies and run diagnostic checks on the generated architectural foundation.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 9, 2026, 01:31 AM