Is design-templates/web-prototype-taste-editorial safe?
https://github.com/nexu-io/open-design/tree/main/design-templates/web-prototype-taste-editorial
This skill is a pure design-guidance document specifying colors, typography, layout rules, and motion constraints for editorial-minimalist web prototypes. The SKILL.md contains no prompt-injection payloads, no file-read or network-exfiltration instructions, no executable code, and no git hooks or submodules. Sensitive file accesses observed in monitoring are attributable to the audit framework's own pre- and post-install canary checks, not to any skill-triggered behavior; all honeypot files remained unmodified. The only noteworthy behavior is the standard monorepo sparse-checkout install pattern and a passive citation of an upstream design repository.
Category Scores
Findings (4)
INFO Monorepo sparse-checkout pattern transiently clones full repo -6 ▶
The install script shallow-clones the entire nexu-io/open-design repository before extracting the target subdirectory. The temporary clone is properly removed. No unexpected network destinations observed.
INFO Canary file reads attributable to audit framework, not skill -3 ▶
Reads of .env, SSH key, AWS credentials, .npmrc, Docker config, and GCP credentials occurred at pre-install and post-install timestamps, consistent with the audit system's own baseline and integrity-check phases. Zero writes to these files. Canary integrity confirmed clean.
INFO Passive external repository citation -3 ▶
The Source section names an upstream GitHub repository. No directive instructs the agent to fetch or visit it.
LOW Skill produces high-fidelity premium-brand UI patterns -5 ▶
Specific color, typography, and layout constraints reliably produce output resembling Notion/Linear-style surfaces. Theoretical downstream phishing surface if misused; no active malicious instruction present.