domain-analysis

Pass

Audited by Gen Agent Trust Hub on Nov 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted business artifacts and discovery notes to derive domain models, exposing a surface for indirect prompt injection.
  • Ingestion points: SKILL.md (Step 1) loads context from paths defined in docs/project-config.json, including team artifacts and discovery notes.
  • Boundary markers: Absent. The instructions do not use delimiters or specific instructions to disregard malicious commands embedded within the artifacts.
  • Capability inventory: The skill uses spawn_agent to create subagents, executes shell commands (git, rg), and performs file read/write operations.
  • Sanitization: Absent. No escaping or validation is performed on the ingested business context before it is used to influence the agent's modeling logic.- [COMMAND_EXECUTION]: The skill utilizes shell commands and local scripts for codebase discovery and analysis.
  • Evidence: mode-review.md executes git diff, git ls-files, and rg (ripgrep) to identify affected files.
  • Evidence: mode-review.md invokes a local Python script located at .claude/scripts/code_graph for structured code analysis.- [DYNAMIC_EXECUTION]: The skill dynamically generates instructions for sub-agents and executes them at runtime.
  • Evidence: mode-review.md (Phase 3) describes constructing a "brief" for a code-reviewer sub-agent based on project conventions and target files.
  • Evidence: The skill utilizes the spawn_agent capability to execute these dynamically generated sub-agent tasks.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 4, 2026, 10:12 AM