What problem does it solve?
It prevents SDD pipeline runs from finishing with unverified feature behavior by running a dedicated feature-level testing stage after implementation, with safe guardrails and clear escalation paths.
Core Features & Use Cases
- Depth-gated feature testing: Prompts the user for quick vs standard testing depth, then reuses that depth consistently across fix-loop iterations.
- Tool-aware tester strategy: Delegates execution to a tester subagent that selects UI/DB/Mixed approaches based on plan type and available tools, and falls back to code review when MCP tools are unavailable.
- Bounded fix loop: Runs a coordinator-driven PASS/FAIL/MCP_UNAVAILABLE workflow with a maximum of 3 fix iterations, escalating to the user if the loop can’t resolve issues.
- Safety and policy enforcement: Includes explicit prohibitions around state handling (never committing
state.json) and forbids the coordinator from self-testing when MCP tools are missing.
Quick Start
Load this skill to test the just-implemented feature by first answering the requested testing depth (quick or standard) when prompted.