tailscale
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFECREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The documentation references a specific SSH user identity.
- Evidence: The file
references/tailscale-troubleshooting.mdincludes an example commandssh kalmyk@temporal-grpc.ide-newton.ts.netwhich hardcodes a specific username. - [INDIRECT_PROMPT_INJECTION]: The skill implements a diagnostic script that processes external hostname strings, creating a potential surface for injection if the agent provides unvalidated input.
- Ingestion points: The
scripts/tailscale-diag.shscript accepts a hostname as its first positional argument ($1). - Boundary markers: There are no delimiters or specific instructions to the agent to treat the input as untrusted or to ignore embedded control characters.
- Capability inventory: The skill executes local CLI commands (
tailscale status,tailscale netcheck, andtailscale ping) via a shell script. - Sanitization: While the script uses standard shell quoting (
"$HOST") to prevent simple word-splitting, it does not perform strict validation or sanitization of the hostname string before execution.
Audit Metadata