fixlisted

Fix issues end-to-end across data pipelines (Airflow/dbt), app stack (backend/frontend), and infra (CI/CD, Terraform, K8s). Scout → diagnose → apply at root cause → verify with fresh evidence → add regression guard. Use for failing DAGs, dbt test failures, 5xx, UI regressions, GH Actions failures, terraform drift, CrashLoopBackOff, lint/type errors. Stops after 3 failed attempts to question architecture.
vanducng/skills · ★ 8 · DevOps & Infrastructure · score 81

Install: claude install-skill vanducng/skills

# Fix End-to-end fixing across the surfaces you actually work on: data pipelines, app stack, infra. Find the cause first, fix at the source, verify with fresh evidence, leave a regression guard so the same class of bug can't return. ## Iron law ``` NO FIX WITHOUT ROOT CAUSE. NO "DONE" WITHOUT FRESH EVIDENCE. ``` Symptom fixes are failure. Random changes waste time and create new bugs. Three failed attempts means the approach is wrong - stop and question architecture, don't keep trying. ## Modes | Mode | When | Behavior | |---|---|---| | **default** | Standard issue, you want it done right | Full loop: scout → diagnose → apply → verify → prevent. Pauses for confirmation if the fix touches >3 files or crosses surfaces. | | `--quick` | Trivial (lint, single type error, obvious typo, known recipe) | Skip deep diagnosis. Still verify with fresh evidence. Still add regression test if behavior changed. | | `--auto` | You trust the loop, end-to-end run | No confirmation gates. Stops only on verification failure or 3rd failed attempt. | | `--no-prevent` | Throwaway / spike / hotfix where guard will land in follow-up | Skip regression-test step. Loud warning. Use sparingly. | Detect mode from the argument; announce in your first reply. ## Workflow 1. **Scout** - locate affected code/models/manifests (`vd:scout` or 2-3 Explore agents) 2. **Diagnose** - activate `vd:debug`; structured root-cause analysis; capture pre-fix evidence 3. **Assess scope** - quick | standard | deep | p