011matthias

build-test-fix

Autonomous build-test-diagnose-fix iteration loop with outcome verification. Use after initial implementation to verify both execution success AND output correctness. Classifies errors and outcome mismatches, applies known fixes, retests, and escalates only for novel problems. Works across Make.com, n8n, and Trigger.dev.

011matthias 0 Updated 5mo ago

Resources

1
GitHub

Install

npx skillscat add 011matthias/agentic-ops1-01/build-test-fix

Install via the SkillsCat registry.

SKILL.md

Build-Test-Fix Loop

Wraps the iterative cycle that happens after every initial build: execute → check results → verify outcomes → classify failure → apply fix → retest. Runs autonomously for up to 3 iterations before escalating.

When to Use

  • After deploying or updating a scenario/workflow — whether it fails OR succeeds on first run
  • When execution errors appear in logs (Make.com execution inspector, n8n execution history)
  • When the build skill completes but the scenario hasn't been verified working
  • After any build where output correctness must be verified (not just execution success)
  • Invoked automatically by the build-orchestrator after implementation phase

Quick Start

  1. Identify the scenario/workflow that needs testing
  2. Run it (or trigger via webhook)
  3. Read execution results
  4. If failure → classify → fix → retest (up to 3 iterations)
  5. If success → report and move on

Escalation Policy

Condition Action
Iterations 1-3 Attempt autonomous fix using FIX-PATTERNS registry
Same category + same fix approach repeated Escalate early — going in circles, need different strategy
3 iterations exhausted without resolution Escalate to user with full diagnosis
Novel error (not in registry) Escalate immediately with context
Critical rule Never retry the same fix twice — each iteration must try a DIFFERENT approach

Orchestrator Dispatch

Orchestrator Execute Read Results Fix
Make.com scenarios_run or webhook POST executions_listexecutions_get_detail Update blueprint via tools/make-api.py update (NOT scenarios_update — returns 500 on blueprints)
n8n MCP workflow execution n8n execution history via MCP Update workflow JSON via MCP
Trigger.dev npx trigger.dev@latest dev + trigger Task run logs Edit TypeScript source

Modules

Module Purpose
ITERATION-LOOP.md Step-by-step autonomous loop procedure
OUTCOME-VERIFICATION.md Verify outputs are correct, not just that execution succeeded
FAILURE-TAXONOMY.md Error classification guide with detection patterns (incl. OUTCOME_MISMATCH)
FIX-PATTERNS.md Registry of known failure→fix mappings
E2E-PIPELINE-VERIFICATION.md Multi-scenario pipeline testing (trigger chain mapping, parallel trigger detection)