ln-520-test-planner
Pass
Audited by Gen Agent Trust Hub on Apr 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Retrieves shared reference markdown files and configurations from the author's official GitHub repository (
github.com/levnikolaevich/claude-code-skills) to supplement local documentation when files are missing.\n- [PROMPT_INJECTION]: The skill ingests data from external project management sources (Linear comments) and local project files (kanban boards) to determine its execution path, which represents a surface for potential indirect prompt injection.\n - Ingestion points: Reads task status and research history from
docs/tasks/kanban_board.mdand external Linear comments.\n - Boundary markers: No delimiters or instructions to ignore embedded commands are implemented for external data inputs.\n
- Capability inventory: Orchestrates sub-agents that have the capability to perform file system operations and execute bash scripts (via
ln-522-manual-tester).\n - Sanitization: External comment content is utilized to verify the state of implementation without explicit validation or filtering steps.
Audit Metadata