Set up Cloudflare Turnstile end-to-end in a project — scan the codebase, create the widget via the Cloudflare API, deploy the managed siteverify Worker, write the frontend snippets, validate, and persist the skill. Load this when a user ask…
What we found · 7
Possible data-exfiltration intentcritical"Send … to <server>" phrasing; combined with secret access this forms a complete data-leak path.
Description doesn’t say WHEN to use itmediumA good description names the trigger (e.g. "when reviewing a PR"). Without it the skill mis-fires or never fires.
Bloated: high token costmedium~3,759 tokens on every single call. Compress instructions and trim examples.
Makes outbound network requestsmediumSends or pulls data from external endpoints — a data-egress channel; confirm the target is trusted.
Contains curl / wget network commandmediumAccesses the network from the shell. If the target is untrusted it can pull malware or exfiltrate data.
Privilege escalation / changes exec permission — in a warninglowAppears inside a prohibition/warning ("never …"), not as an instruction to run. Flagged for awareness only.
Hard-coded IP addresslowLiteral IPs are often used to bypass domain review for egress — worth verifying.
What its scripts can do
medium
Network egress
Conflicts with · 8
cloudflare-email-service
shadow
cloudflare, worker, api
internal-comms
shadow
set, write
docx
shadow
create
wrangler
overlap
cloudflare, deploy, manag, worker, load
agents-sdk
overlap
cloudflare, worker, load
workers-best-practices
overlap
cloudflare, worker, load
cloudflare
overlap
cloudflare, worker
durable-objects
overlap
create, cloudflare
If you have both installed, they compete for the same triggers — the top cause of an agent picking the wrong skill.
Have you installed turnstile-spin? Scan everything you run.
One command grades every skill in your Claude Code / Codex — safety, token bloat, and conflicts, 100% locally.
Does this rating match your experience with turnstile-spin?
Anonymous & aggregate. We watch these as a trend and re-check flagged skills — a single click never moves a grade on its own.
Independent static analysis by SkillMOO — the Consumer Reports for agent skills. Safety, token, and conflict grades are computed, not opinions. Efficacy is only ever claimed when measured. We rate & link to the source; we do not rehost skill content.