kmp-expert
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill performs local script execution and build tool invocations as part of its orchestration and validation workflows.
- Invokes
python3 skills/kmp-expert/scripts/validate_skill_map.pyto ensure consistency in the skill registry. - Executes
git cloneto initialize new projects from established templates. - Uses
./gradlewcommands for build, test, and verification tasks in generated projects. - [EXTERNAL_DOWNLOADS]: Downloads project templates and references libraries from trusted or well-known sources.
- Fetches the
multiplatform-library-templatefrom the official JetBrainsKotlinorganization on GitHub. - References Maven dependencies such as
io.github.ronjunevaldoz:shadcn-compose, which are maintained by the skill author. - [INDIRECT_PROMPT_INJECTION]: The skill generates architectural plans and agent configuration based on user-provided project descriptions, creating a surface for indirect prompt injection.
- Ingestion points: Project descriptions and intake parameters processed in
references/new-project-phase-1-intake-and-plan.md. - Boundary markers: Implements a multi-step planning process with mandatory user confirmation gates (e.g., Steps 3c, 8a, 8c) before code generation or execution.
- Capability inventory: Includes file system modification, project scaffolding, and shell command execution capabilities via Gradle and Python scripts.
- Sanitization: Instructions do not define explicit sanitization or escaping for user-provided descriptions before they are used to generate planning documents or configuration files.
Audit Metadata