030-architecture-adr-general
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill workflow includes executing
./mvnw validateormvn validateto ensure project health before documenting decisions. It also uses thedatecommand to retrieve the current system time for document headers. These are legitimate operations for a Java development assistant and do not involve user-controlled input in the command strings. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided responses during a conversational interaction to generate architectural documentation.
- Ingestion points: User input gathered from a series of 15 questions regarding architectural problems, stakeholders, and chosen options defined in
references/030-architecture-adr-general.md. - Boundary markers: The skill uses a structured conversational flow and a static markdown template to encapsulate user input, reducing the risk of accidental instruction obedience.
- Capability inventory: The skill can execute local build commands (
mvn), system utilities (date), and perform file system writes to create markdown files. - Sanitization: User input is primarily used for descriptive text in markdown files. The skill applies kebab-case transformation to user-provided titles when generating filenames, which mitigates basic command injection risks in the file system.
Audit Metadata