architect
Pass
Audited by Gen Agent Trust Hub on Aug 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes Git commands to ensure the project is up-to-date. Evidence:
git fetch,git rev-parse, andgit rev-listinSKILL.md. - [EXTERNAL_DOWNLOADS]: The skill can install new agent capabilities. Evidence:
npx skills add <owner>/<repo>ininternal/tool-discovery.md. This is gated by a mandatory user consent panel. - [EXTERNAL_DOWNLOADS]: The skill performs web research using a dedicated
researchersubagent to fetch technical documentation. - [PROMPT_INJECTION]: The skill processes external data, creating a surface for indirect injection. Ingestion points: Local codebase,
AGENTS.md, and web summaries. Boundary markers: Uses structured markdown templates. Capability inventory: File writing and bash execution. Sanitization: Uses restricted subagents and specific delimiters.
Audit Metadata