dev-plan
Pass
Audited by Gen Agent Trust Hub on Jun 21, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill incorporates a surface for indirect prompt injection.
- Ingestion points: It gathers context from files within
.harness/product/,README.md, andAGENTS.md. - Boundary markers: No explicit delimiters or instructions to ignore embedded commands are used when passing this context to subagents.
- Capability inventory: The skill can write and update multiple markdown files and documentation across the repository and suggests the execution of external tools.
- Sanitization: There is no evidence of validation or filtering for the data retrieved from the local repository before it is processed by subagents.
- [COMMAND_EXECUTION]: The skill includes instructions to validate design tokens using the command
npx @google/design.md@0.2.0 lint DESIGN.md. This involves executing a CLI tool from a known provider on a generated file. - [EXTERNAL_DOWNLOADS]: The skill references the
@google/design.mdpackage from the NPM registry and links to an official documentation repository on GitHub. Both sources are associated with a well-known technology service provider.
Audit Metadata