TDD Orchestrator with Parallel Execution

VerifiedCaution

Executes the full TDD workflow: reads a ticket and associated knowledge, then runs TDD tests in parallel for all ready implementation tasks. After parallel execution, it launches a refine loop to fix any code-test issues. Ideal for accelerating TDD cycles when multiple independent test implementations are available.

Sby Skills Guide Bot
TestingAdvanced
1706/2/2026
Claude Code
#tdd#test-driven-development#parallel-testing#workflow#orchestration

Recommended for

Our review

Automates the full test-driven development workflow by orchestrating parallel test execution and refinement loops.

Strengths

  • Parallel execution of multiple TDD tasks for faster feedback
  • Seamless integration with existing ticket and design documentation
  • Automatic refinement loop to improve code and tests iteratively
  • Comprehensive report generation summarizing results

Limitations

  • Requires a pre-existing design.md file for the ticket
  • Depends on external scripts and the Claude Code environment
  • Overhead may be excessive for very simple or small tasks
When to use it

Use when you have a well-defined ticket with a design document and need to implement and test multiple components concurrently with automatic refinement.

When not to use it

Avoid for trivial tasks where the orchestration overhead outweighs the benefit, or when the design is not yet documented.

Security analysis

Caution
Quality score90/100

The skill uses Bash to run scripts that manage dependencies and trigger hooks. This is a legitimate use but involves arbitrary command execution within the agent environment, warranting caution.

Findings
  • Executes bash commands to run internal Node scripts (resolve-task-dependencies.js and task-runner-hook.js). While these appear to be internal tools, their content is not reviewed, and misuse could lead to arbitrary code execution.
  • Uses the Bash tool extensively, which can modify files and interact with the system, posing a risk if the invoked scripts are compromised or malformed.

Examples

Basic TDD orchestration
/tdd ticket=TICKET-123
TDD with custom iterations
/tdd ticket=TICKET-456 iterations=5
Full workflow with refine loop
/tdd ticket=TICKET-789 iterations=10

name: tdd description: "Executes complete TDD workflow: parallel TDD → refine-loop" user-invocable: true allowed-tools: Read, Write, Edit, Bash, Task, Skill, Glob, Grep, TaskList, TaskGet, TaskUpdate, TaskOutput hooks: Stop: - hooks: - type: command command: "bash -c 'node ~/.claude/scripts/task-runner-hook.js'"

/tdd

Orchestrator for TDD workflow with parallel execution.

Progress Checklist

- [ ] Step 1: Parse args and set variables
- [ ] Step 2: Invoke ticket-reader agent
- [ ] Step 3: Validate design exists
- [ ] Step 4: Invoke knowledge-reader agent
- [ ] Step 5: Execute TDD with parallel orchestration
- [ ] Step 6: Invoke /refine-loop (if no errors)
- [ ] Step 7: Generate report

Steps

  1. Parse args and set variables:

    • ticket → TICKET_ID (required, error if not provided)
    • iterations → MAX_ITERATIONS (default: 10)
  2. Invoke ticket-reader agent:

    • Task prompt: OPERATION=get TICKET_ID=$TICKET_ID
    • If error → END
    • Parse output → TICKET_PATH
  3. Validate design exists:

    • Check if <TICKET_PATH>/design.md exists
    • If not → Error: "design.md not found. Run /design first.", END
  4. Invoke knowledge-reader agent:

    • Task prompt: OPERATION=resolve TICKET_PATH=$TICKET_PATH WORKFLOW=/tdd
    • Parse output → KNOWLEDGE (empty array if error)
  5. Execute TDD with parallel orchestration:

    a. Get ready tasks:

    • Run node ~/.claude/scripts/resolve-task-dependencies.js --ticket=$TICKET_ID --filter="Implement:"
    • If no tasks → Error: "No implementation tasks found", END

    b. For each ready task:

    • TaskGet(task_id) to get task details
    • TaskUpdate(task_id, status="in_progress")
    • Launch tdd-runner with Task tool:
      • subagent_type: tdd-runner
      • run_in_background: true
      • prompt includes: TASK_SUBJECT, TASK_DESCRIPTION, KNOWLEDGE, CWD

    c. Monitor and iterate:

    • Use TaskOutput to check completion
    • On completion: TaskUpdate(task_id, status="completed")
    • Refresh dependencies, launch newly ready tasks
    • Repeat until all complete

    d. Collect results: TDD_FILES, TDD_TEST_COUNT, TDD_PASSED_COUNT

  6. Invoke /refine-loop (if no TDD errors):

    • Args: "type=code-tests ticket={TICKET_PATH} scope=uncommitted iterations={MAX_ITERATIONS}"
  7. Generate report:

    • Write to <TICKET_PATH>/tdd-report.md
Related skills