tommymorgan:work

Execute Plan Autonomously with Quality Gates

Work through all scenarios in a plan using TDD with comprehensive quality gates. Continue autonomously until complete or blocked.

Core Principles

• Bulldog Persistence: Don't give up or take shortcuts that create tech debt
• Border Collie Intelligence: Apply excellent judgment and autonomy
• Root-Cause Driven: Use /tommymorgan:root-cause instead of speculation
• Quality Gates: Code review and exploratory testing before completion
• Documentation First-Class: Update docs alongside code

Completion Criteria

The work command ONLY stops when ALL FOUR conditions are met:

1. All scenarios DONE: Every scenario in plan marked <!-- DONE -->
2. All tests passing: Full test suite passes in local development
3. All layers functional: Every application layer validated and working
4. All living specs written: Every scenario has corresponding .feature file in features/

Before ANY potential stopping point, check completion criteria:

function checkCompletionCriteria():
  scenarios_done = all scenarios marked DONE in plan
  tests_passing = test suite exits 0 in local dev
  layers_functional = all detected layers validated successfully
  living_specs_written = features/*.feature contains all scenarios (verify with grep)

  return scenarios_done AND tests_passing AND layers_functional AND living_specs_written

Verify living specs exist:

# Count scenarios in plan
plan_scenarios=$(grep -c "^Scenario:" plan.md)

# Count scenarios in feature files
feature_scenarios=$(grep -c "^  Scenario:" features/*.feature 2>/dev/null || echo 0)

# Must match (or feature_scenarios >= plan_scenarios)

If checkCompletionCriteria() returns false → continue execution If checkCompletionCriteria() returns true → report completion and stop

NEVER:

• Ask "should I continue?"
• Mention token usage as stopping reason
• Ask "what should I do next?"
• Say "this is separate work"
• Rationalize incomplete work as "done"