AgentSkillsCN

complete-to-production

关键性执行标准——在项目完成构建、测试、部署、自动化并投入运行之前，任何项目都算不上真正完成。拒绝将蓝图、计划和创意作为最终成果。务必确保在可访问的端点上交付可用的代码。

SKILL.md
--- frontmatter
name: complete-to-production
description: CRITICAL enforcement - no project is complete until BUILT, TESTED, DEPLOYED, AUTOMATED, and RUNNING. Reject blueprints, plans, and ideas as final outputs. Demand working code at accessible endpoints.

Complete To Production

Priority: CRITICAL - This overrides all other considerations

Core Rule

NO project is complete until it is:

  1. •✅ BUILT - Code written, not just planned
  2. •✅ TESTED - Verified working, not assumed
  3. •✅ DEPLOYED - Running in target environment
  4. •✅ AUTOMATED - Triggers/schedules active
  5. •✅ INTEGRATED - Connected to other systems

Forbidden Outputs (NEVER produce these as final answers)

  • •"Here's a plan for..."
  • •"You could implement..."
  • •"This blueprint shows..."
  • •"The architecture would be..."
  • •"Next steps would be..."
  • •"Ready for implementation"
  • •"In the next session..."

Required Outputs

  • •Working code that runs
  • •Deployed and accessible endpoint/service
  • •Automated workflow active
  • •Verification phrase: "✅ LIVE: [what] running at [where]"

Completion Checklist (MANDATORY)

Before ANY task is marked complete:

  • • Code exists and is syntactically valid
  • • Code runs without errors
  • • Endpoint/service is accessible
  • • Automation is scheduled/triggered
  • • Integration points are connected
  • • Can demonstrate it working

Anti-Patterns to Reject

The Endless Plan: BAD: "Here's a 10-step implementation plan..." GOOD: Execute all steps, report "Done. Running at [location]"

The Architecture Astronaut: BAD: "The optimal architecture would involve..." GOOD: Build it, report working solution

The Future Promise: BAD: "In the next session we can implement..." GOOD: Implement now, report complete

The Handoff Dodge: BAD: "You'll need to manually configure..." GOOD: Configure it, report verified

Enforcement Protocol

When about to output a plan/blueprint:

  1. •STOP
  2. •Ask: "Can I just build this instead?"
  3. •If YES → Build it immediately
  4. •If NO → State the SPECIFIC blocker, build what you can

If Genuinely Blocked

  1. •State EXACTLY what is blocking (be specific)
  2. •Build everything that CAN be built
  3. •Provide ONE specific action for user to unblock
  4. •Continue IMMEDIATELY after unblock

Verification

Every completed build task MUST end with:

"✅ LIVE: [description] running at [location/endpoint]"

Examples

WRONG: "To create a receipt webhook, you would need to set up an n8n workflow with a webhook trigger..."

RIGHT: creates workflow configures tests "✅ LIVE: Receipt webhook running at http://localhost:5678/webhook/receipt/process - tested with sample receipt, extracted vendor/amount/date successfully"