temps-design-system
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references supplemental design guidelines hosted on GitHub (jakubkrehel/skills and emilkowalski/skills). These references are informational in nature, intended for the agent to read for stylistic guidance rather than for automated code execution.
- [COMMAND_EXECUTION]: The skill defines several commands to be executed in the development environment using
bun, such asbun run lint,bun run tokens:build, andbun run audit:records. These are standard tasks for a design system's token pipeline and style auditing within a local repository. - [INDIRECT_PROMPT_INJECTION]: The skill provides procedures for auditing a local console UI using an
agent-browseradapter. This creates an interface where the agent processes external content (computed styles and HTML structures). The risk is mitigated by explicit instructions to avoid capturing credentials, input values, or customer records during these audits. - Ingestion points: Browser-rendered routes audited via the
agent-browseradapter. - Boundary markers: Not explicitly defined for the data stream, but the skill provides high-level scope boundaries.
- Capability inventory: Local file reads (DESIGN.md, RULES.md), shell command execution via
bun, and browser interaction viaagent-browser. - Sanitization: Explicitly instructs the agent to exclude credentials, input values, and customer records from its audit findings.
Audit Metadata