kata-write-skill
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill acts as a set of authoring guidelines for creating other agent skills. It does not include any malicious commands, hidden payloads, or suspicious network operations.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external sources or existing skill packages (ingestion points in SKILL.md and references/import.md). It mitigates potential injection risks by mandating a security 'preflight' check that specifically inventories scripts, network destinations, and instructions that override user rules (boundary markers and sanitization). Execution of imported code is restricted to isolated temporary workspaces without user secrets (capability management).
- [DYNAMIC_EXECUTION]: The skill describes routines for creating verification packages that launch and drive project-specific commands (references/create-verification-skill.md). These are legitimate development workflows and include instructions for process teardown and data isolation.
Audit Metadata