Skip to content

support-and-scale-debug-errors-guide.md

Latest commit

 

History

History
124 lines (92 loc) · 3.66 KB

support-and-scale-debug-errors-guide.md

File metadata and controls

124 lines (92 loc) · 3.66 KB

Support & Scale: Debug Errors with Amp

Use this prompt‑first workflow to triage incidents quickly, trace failures precisely, implement minimal‑risk fixes, and validate outcomes.

Note: Run Amp from your repo root (CLI) or open the workspace in your IDE; there’s no need to include repository paths in prompts.

Why this phase matters

  • Restores stability and meets SLAs/SLOs faster.
  • Prevents repeat failures with targeted tests and observability.
  • Turns firefighting into a repeatable practice.
  • Captures knowledge for future responders.

Pre‑flight checklist

  • Repro: failing input, expected vs actual behavior, impact scope.
  • Context: version/commit, environment, feature flags, recent changes, owners.
  • Access: code, logs, traces, dashboards, error pages.
  • Safety: separate branch and rollback/flag plan.

1) Map the repo (error surfaces)

Prompt (single message):

Inventory logging, metrics, tracing, and error boundaries for the suspected path. Note entry points,
cross‑service calls, and highest‑risk choke points with blast radius.

What you’ll get:

  • Error surface map for the failing area
  • Cross‑service and choke‑point awareness
  • Immediate high‑signal probes to add

2) Read the history (incidents, regressions)

Prompt (single message):

Summarize related incidents, recent regressions, risky diffs, and flaky tests. Cluster similar stack traces
and suggest likely defect classes.

What you’ll get:

  • Incident and regression context
  • Risky pattern inventory with hypotheses
  • Reviewer/owner suggestions

3) Read with intention (trace the failing flow)

Prompt (single message):

Trace the failing flow end‑to‑end and include file:line pointers for each hop. Highlight guardrails,
invariants, and missing checks. Propose high‑signal log/tracing additions.

What you’ll get:

  • File:line trace of the failure path
  • Guardrails and missing checks identified
  • Observability improvements to land

4) Implement (minimal‑risk fix)

Prompt (single message):

Propose the smallest safe change to prevent this failure. Include toggles/flags, blast radius,
side effects, and a rollback plan. Provide a focused diff with rationale.

What you’ll get:

  • Minimal fix diff and rationale
  • Flag/rollback guidance and risk notes
  • Side‑effect review

5) Validate (repro → tests → checks)

Prompt (single message):

Reproduce the failure and confirm it disappears. Add/update unit/integration tests for the defect shape.
Run targeted deployment checks (logs quieting, error rates normalizing, user path restored).

What you’ll get:

  • Repro evidence pre/post fix
  • Tests guarding against recurrence
  • Deployment health confirmation

Exit criteria and handoffs

  • Root cause documented with precise trigger and scope
  • Minimal fix merged with tests and observability added
  • Error/latency rates normal; no new warnings in adjacent areas
  • Postmortem/runbook updated; follow‑ups created and owned

Quick reference

Prompt templates

Map error surfaces for <path>; list choke points and probes.
Cluster incidents/regressions and propose likely defect classes.
Trace failure with file:line pointers and missing checks.
Minimal fix with flags/rollback and focused diff.
Verify: repro disappears, tests added, deployment checks green.

Developer‑run verification (optional)

git switch -c fix/debug-errors && git status
npm test -- --run --no-color || pnpm test -- --run --no-color

Closing note

Use Amp as your debugging copilot: ask for maps, traces with file:line pointers, small fixes, and crisp validation. Amp returns synthesized results—not raw logs.