architecture-model-bootstrap

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill's design focuses on architectural modeling with explicit constraints to keep proposals grounded in existing worktree evidence and manifests.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests potentially untrusted data from external sources within the project worktree.
  • Ingestion points: Project manifests (package.json, Cargo.toml, etc.), README.md, AGENTS.md, and the docs/ directory as listed in the evidence pack section of SKILL.md.
  • Boundary markers: The skill includes 'Grounding rules' that prohibit inventing structure without code evidence and specifies that documentation is used for naming/ownership only, not for structure the code contradicts.
  • Capability inventory: No capabilities for network access, file writing (unless manually requested), or arbitrary command execution were identified; the skill produces a structured JSON proposal.
  • Sanitization: Validation against a predefined schema in references/model-schema.md is required before outputting the model.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 01:26 PM
Security Audit — agent-trust-hub — architecture-model-bootstrap