create-subagent
Pass
Audited by Gen Agent Trust Hub on Mar 18, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The subagent templates in SKILL.md are vulnerable to indirect prompt injection.
- Ingestion points: The 'Code Reviewer' and 'Debugger' templates ingest untrusted data from git diff, modified source files, and system error logs.
- Boundary markers: The provided templates do not include delimiters or instructions to ignore embedded prompts in the analyzed data.
- Capability inventory: The templates encourage subagents to read files and execute CLI tools such as git and bq (BigQuery).
- Sanitization: There is no provision for sanitizing or validating external content before it enters the subagent's context.
Audit Metadata