Is shadcn/ui safe?
Yes. shadcn/ui is safe to install. Oathe's behavioral security audit gave the ui skill by shadcn a trust score of 96/100 with 4 findings, none critical or high.
Is shadcn/ui safe to install?
The shadcn/ui repository is the official, well-established open-source UI component library maintained by shadcn. Both SKILL.md files found (migrate-radix-to-base and shadcn) are well-structured, appropriately scoped, and free of prompt injection, obfuscation, or malicious directives. Sensitive file accesses observed during monitoring are attributable to the audit framework's own canary initialization rather than repository code, as confirmed by the passing canary integrity check and the absence of any external data transmission to unexpected hosts.
What security issues were found in shadcn/ui?
Category Scores
Findings (4)
LOW Canary files opened during monitoring initialization -7 ▶
Filesystem monitoring captured opens of .env, .ssh/id_rsa, .aws/credentials, .npmrc, .docker/config.json, and GCloud credentials at 07:36:39. The timing (at session start, before any git/find operations) and the passing canary integrity check strongly indicate these accesses are from the audit framework's own canary setup routine, not from repository code. No data was sent externally and all files remained unmodified.
Evidence: 07:36:39 OPEN /home/oc-exec/.env; 07:36:39 OPEN /home/oc-exec/.ssh/id_rsa; 07:36:39 OPEN /home/oc-exec/.aws/credentials — canary integrity check: ✅ All canary files intact
INFO Shadcn skill auto-activates via user-invocable: false -3 ▶
The skills/shadcn/SKILL.md declares user-invocable: false, meaning it activates automatically when a user is working in a project with components.json. This expands agent permissions (shadcn CLI, file reads) without explicit per-task user consent. The scope is well-bounded to shadcn-related commands and is consistent with the skill's documented purpose.
Evidence: skills/shadcn/SKILL.md: 'user-invocable: false' and 'allowed-tools: Bash(npx shadcn@latest *), Bash(pnpm dlx shadcn@latest *), Bash(bunx --bun shadcn@latest *)'
INFO Repository was pre-existing; no fresh clone executed 0 ▶
The install step returned a fatal error indicating the repository directory already existed. All file access and network activity observed during the audit session reflects the monitoring framework's inspection of a pre-existing checkout, not a fresh installation. This is neutral but noted for completeness.
Evidence: Install output: 'fatal: destination path '/home/oc-exec/skill-under-test/shadcn/ui' already exists and is not an empty directory.'
INFO Skills instruct agent to execute shadcn CLI commands -3 ▶
Both SKILL.md files direct the agent to run npx/pnpm/bun shadcn@latest commands as part of normal operation. This is the declared purpose of the skills and the commands are appropriately restricted via allowed-tools. No install hooks, git hooks, or remote code download patterns were found in repository scripts.
Evidence: skills/shadcn/SKILL.md allowed-tools; root package.json scripts; workspace package.json files — no preinstall/postinstall hooks present
Should I install shadcn/ui?
Oathe's verdict for shadcn/ui is SAFE with a trust score of 96/100. Recommendation: Install.