skills/lassejlv/skills/paper-to-gpui/Gen Agent Trust Hub

paper-to-gpui

Pass

Audited by Gen Agent Trust Hub on Aug 16, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes a local bash script scripts/inspect_gpui_project.sh to perform a read-only inventory of the target Rust project. This script uses standard system utilities such as git, find, grep, and awk to identify dependencies and source files without performing any destructive or network-facing operations.
  • [COMMAND_EXECUTION]: The workflow includes standard Rust development commands such as cargo check, cargo test, cargo clippy, and cargo fmt. These are executed within the user's local repository to validate code correctness and visual fidelity, which is consistent with the skill's primary purpose.
  • [EXTERNAL_DOWNLOADS]: The skill interacts with a local Paper Desktop MCP server at http://127.0.0.1:29979/mcp. This connection is restricted to the local loopback address, ensuring that design data remains local to the user's machine.
  • [INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface for indirect prompt injection by ingesting data from Paper designs via the MCP server (e.g., get_jsx, get_computed_styles). However, the instructions explicitly command the agent to treat this data as a 'structural hint' rather than 'unquestioned truth' and require the agent to perform visual validation against a screenshot. The execution of project-local code via cargo test is a standard development capability. Boundary markers are present in the documentation directing the agent to establish context and verify information before implementation.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 16, 2026, 12:59 AM
Security Audit — agent-trust-hub — paper-to-gpui