claude-content-writer: why it is not safe to use

What we found

Our review did not find hidden instructions, hardcoded secrets, credential access, privilege escalation, or evidence that this repo itself sends user content to an undisclosed party. The main concern is the install behavior. The package's postinstall flow automatically runs a shell installer and clones additional skills without an explicit confirmation for each action. During install, it fetches third-party code from named GitHub repositories for `claude-seo` and `humanizer`, and it may also run `claude-seo`'s `install.sh` if present. That increases trust placed in code outside the primary repo at install time and can make it harder to review exactly what will execute before setup completes.

What to do instead

Prefer skills that keep installation self-contained, or that require a clear prompt before running shell scripts or pulling extra repositories. If you still want to use this skill, review the postinstall logic first, inspect the referenced GitHub repos and any `install.sh` scripts, and pin versions or commit hashes where possible. Running setup in an isolated environment can also reduce risk while you verify what additional code is fetched and executed.