wireframe-skill
Warn
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_CONTEXT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructs the agent to execute installation commands (
npx skills add yhassy/wireframe-skill) which downloads and installs code from an external GitHub repository. This effectively allows the execution of unverified remote code on the user's system. - [DYNAMIC_CONTEXT_INJECTION]: The
SKILL.mdfile utilizes the!command syntax to execute shell scripts at load time. These scripts check for existing file paths and grep for specific strings in~/.agents/skills/wireframe/SKILL.mdto determine installation status. While used for logic gating, this mechanism executes shell code automatically upon loading the skill. - [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it is designed to ingest data from untrusted external sources.
- Ingestion points: Reads layout data from
designs/{design-id}/output/design-spec.md. - Boundary markers: No explicit delimiters or instructions to ignore embedded commands are mentioned for the design spec processing.
- Capability inventory: The skill executes Python 3 one-liners, generates interactive HTML files, and writes machine-readable JSON to the filesystem.
- Sanitization: No sanitization or validation of the input design specification is described.
- [EXTERNAL_DOWNLOADS]: The skill facilitates the download of a full design tool from a third-party GitHub repository (
github.com/yhassy/wireframe-skill) during the installation phase.
Audit Metadata