stelow-workflow-tech-planning
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external product specifications and user input to drive the planning process, creating a surface for potential injection.\n
- Ingestion points: The skill reads product specification files (e.g.,
spec-product*.md) and user input obtained through theask_user_questiontool.\n - Boundary markers: It includes a mandatory security check verifying an
approved: trueflag in the specification metadata and uses acontext: "fresh"directive for its planner subagent.\n - Capability inventory: The skill performs file operations, executes shell commands for stack detection and codebase analysis, and manages project goals.\n
- Sanitization: There is no evidence of explicit sanitization or escaping of the ingested markdown content before it is processed by the agent or delegated to subagents.\n- [EXTERNAL_DOWNLOADS]: The skill leverages
npxto download and execute external packages at runtime.\n - Fetches the
@vedanth/context7package for documentation searching.\n - Downloads the
skillspackage from Vercel Labs to access tech-stack-specific prompt templates.\n- [COMMAND_EXECUTION]: The skill contains multiple shell scripts designed to be executed by the agent to detect the project stack, perform codebase reconnaissance, and validate planning artifacts.
Audit Metadata