kernel-ship

Pass

Audited by Gen Agent Trust Hub on Jul 11, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill relies on external repository content, such as runbooks and deployment scripts, which introduces an indirect prompt injection surface. A compromised repository file could potentially influence the commands executed by the agent.\n
  • Ingestion points: Ingests target deployment scope and environment-specific runbooks (SKILL.md).\n
  • Boundary markers: Includes a manual user confirmation step ("Ship now?") before proceeding to execution.\n
  • Capability inventory: Authorized to use the bash tool for executing deployment commands.\n
  • Sanitization: Relies on the user's manual approval gate but does not perform automated sanitization of the documentation-derived commands.\n- [COMMAND_EXECUTION]: The skill is designed to perform deployment tasks using the bash tool. It includes guardrails instructing the agent to use only documented project commands and runbooks rather than inventing deployment steps, which mitigates the risk of non-standard or hallucinated command execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 11, 2026, 09:05 AM
Security Audit — agent-trust-hub — kernel-ship