architecture-decisions
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external data during the RESEARCH phase by reading source code and documentation of dependencies. \n
- Ingestion points: The agent reads external project code and documentation to summarize findings (references/adr-lifecycle.md). \n
- Boundary markers: The skill explicitly requires creating docs/decisions/CLAUDE.md and docs/decisions/AGENTS.md to instruct agents to ignore ADR history for general implementation guidance. \n
- Capability inventory: The skill uses git for branching and commits, and gh for creating pull requests (references/adr-template.md). \n
- Sanitization: Relies on a four-phase lifecycle (RESEARCH, DRAFT, HOLD, MERGE) where human review and explicit authorization are mandatory gates. \n- [COMMAND_EXECUTION]: The skill uses standard development environment tools for workflow management. \n
- Evidence: Executes git for repository management and gh for interaction with GitHub Pull Requests (references/adr-template.md). \n- [EXTERNAL_DOWNLOADS]: The skill references an installation command for additional related skills from the same author. \n
- Evidence: Mentions npx skills add jwilger/agent-skills --skill event-modeling to fetch related tooling from the author's repository (SKILL.md).
Audit Metadata