pp-gisis

Pass

Audited by Gen Agent Trust Hub on Jul 20, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTIONCREDENTIALS_UNSAFECOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches binaries from the author's infrastructure using npx -y @mvanhorn/printing-press-library and go install github.com/mvanhorn/printing-press-library/....
  • [DATA_EXFILTRATION]: The CLI supports a --deliver webhook:<url> flag, which enables the redirection of vessel particulars and registry data to arbitrary external HTTP endpoints.
  • [DATA_EXFILTRATION]: A feedback mechanism is present that can be configured via environment variables (GISIS_FEEDBACK_ENDPOINT) to automatically transmit local data to a remote server.
  • [CREDENTIALS_UNSAFE]: The authentication flow depends on a companion tool (press-auth) that captures session cookies from a browser and stores them in the macOS keychain for the CLI to access.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection.
  • Ingestion points: Ship data is retrieved from the external gisis.imo.org registry via ship get and ship batch commands.
  • Boundary markers: The instructions lack explicit delimiters or warnings to ignore instructions embedded within the retrieved maritime data.
  • Capability inventory: The skill uses the Bash tool, which allows for shell command execution.
  • Sanitization: There is no mention of sanitization or validation of external data before it is introduced into the agent's context.
  • [COMMAND_EXECUTION]: The skill is designed to execute the gisis-pp-cli binary through the Bash tool to perform vessel lookups and state management.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 20, 2026, 11:00 PM
Security Audit — agent-trust-hub — pp-gisis