Implement GPT Operativo Completo Plan

Purpose

Execute the GPT Operativo Completo -- Plan de Ejecucion safely and completely, including:

• backend JSON-error hardening
• OpenAPI spec hardening
• GPT instruction system completion
• go-live operational docs

Inputs Expected

• Plan file path (commonly in .cursor/plans/ or attached by user)
• Repositories involved:
  • app repo (Calculadora-BMC)
  • GPT config repo (Calculadora-BMC-GPT)

If the user provides both repos implicitly in prior context, proceed without extra questions.

Mandatory Execution Rules

1. Read the plan first and mirror its sections 1:1.
2. Do not edit the plan file itself.
3. If to-dos already exist, do not recreate them.
4. Set only one todo to in_progress at a time, in order.
5. Mark todo completed only after edits and basic verification are done.
6. Do not stop at partial completion; finish all to-dos unless blocked.
7. Keep responses and docs in professional Spanish unless user requests otherwise.

Workflow

Copy this checklist and keep it updated while executing:

Plan Execution Progress:
- [ ] 1. Backend hardening
- [ ] 2. OpenAPI specs hardening
- [ ] 3. GPT instructions completion
- [ ] 4. Go-live documentation completion
- [ ] 5. Lint/consistency verification
- [ ] 6. Final delivery summary

1) Backend hardening

Target files:

• server/index.js
• server/routes/calc.js

Required outcomes:

• Global cors() active.
• 404 returns JSON (never HTML default page).
• calc handlers return JSON on failure (try/catch where needed).

2) OpenAPI specs hardening

Target files:

• gpt/actions/openapi.yaml
• gpt/actions/openapi-wolf.yaml
• gpt/actions/openapi-mercadolibre.yaml

Required outcomes:

• Single production server entry per spec.
• Stable operationId names.
• Explicit JSON error schemas (400/401/500 as applicable).
• Unified scenario enums (camara_frig, not camara_frigorifica).
• No broken refs.

3) GPT instructions completion

Target files:

• gpt/instructions/system.md
• gpt/instructions/session-qualification.md
• gpt/instructions/manual-gates.md (create if missing)
• gpt/instructions/maintenance.md

Required outcomes:

• Explicit decision tree: intake -> qualify -> quote -> persist -> approve.
• Tool-safety and error-handling policy.
• Manual confirmation contract with CONFIRMATION::....
• Versioning and operationId governance guidance.

4) Go-live docs completion

Target files:

• docs/builder-setup-checklist.md (create/update)
• gpt/examples/payload-reference.md (create/update)
• docs/autopilot-agent.md (update)

Required outcomes:

• Upload steps for all actions.
• Expected operations table.
• Smoke-test payloads.
• Troubleshooting for Failed to Parse JSON.

5) Verification

Minimum verification steps:

• Re-read edited files for syntax/consistency.
• Run lints on changed files where available.
• Confirm no placeholder production URLs remain unless intentionally documented as placeholders.

Final Output Format

Return in this structure:

1. What was implemented (by plan section)
2. Files changed (grouped by repo)
3. Verification results
4. Remaining risks / follow-up actions

Keep it concise and operational.

Additional Resources

• Detailed implementation guardrails: reference.md
• Ready-to-use execution examples: examples.md