turnstile-spin

Pass

Audited by Gen Agent Trust Hub on Oct 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes several local shell scripts provided within the skill package, such as auth-probe.sh, widget-create.sh, worker-deploy.sh, validate.sh, and persist-skill.sh. It also invokes the wrangler CLI tool using npx to manage Workers and secrets.
  • [EXTERNAL_DOWNLOADS]: The instructions guide the user to install the wrangler CLI tool via npm, which is the official tool provided by Cloudflare for interacting with their platform.
  • [DATA_EXFILTRATION]: The skill transmits configuration data and Turnstile metadata to Cloudflare's official API endpoints (api.cloudflare.com). These network operations are necessary for the skill's primary function and target a well-known, trusted service.
  • [CREDENTIALS_UNSAFE]: The skill manages sensitive Cloudflare API tokens. It recommends best practices for handling these, such as using environment variables or saving the token to a file with restricted permissions (umask 077 && printf '%s' '<token>' > ~/.cf-turnstile-token). It explicitly instructs the agent not to write secrets to disk or expose them in conversation logs.
  • [INDIRECT_PROMPT_INJECTION]: The skill performs a scan of the local codebase (e.g., package.json, README.md, source code files) to detect frameworks and insertion points for the Turnstile widget. This represents an ingestion point for untrusted data, though the risk is mitigated by the requirement for the agent to show diffs and wait for user confirmation before applying any changes.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 6, 2026, 04:22 PM