AGENT SKILL

trove-debug

Systematic debugging workflow. Use for bug reports, failing tests, broken behavior, red CI, regressions, or unexplained production symptoms.

License 许可
MIT
所属仓库
bravew/trove
最近更新
2026-11-12
Stars
★ 10
Forks
1

01 前言 Front Matter

name
trove-debug
metadata
{ "trove-category": "engineering", "trove-category-label": "Engineering" }
description
Systematic debugging workflow. Use for bug reports, failing tests, broken behavior, red CI, regressions, or unexplained production symptoms.
when_to_use
bug report; test failure; this is broken; red ci; debug
user-invocable
true

03 SKILL.md 正文

<!-- AUTO-GENERATED from SKILL.md.tmpl — do not edit directly --> <!-- Regenerate: bun run build:skills -->

Session Init

This skill ships Trove conventions. Prefer existing project patterns over generic best practices when they conflict.

If a sibling skill in this plugin matches the request more directly, defer to it. See AGENTS.md (or docs/routing.md in the marketplace) for the per-plugin routing index.

trove-debug

Reproduce, narrow, hypothesize, verify - in that order.

Workflow

  1. Capture the observed failure and expected behavior.
  2. Reproduce it with the smallest command, test, fixture, or manual path available.
  3. Narrow the failing layer by reading code and adding temporary inspection only when needed.
  4. State one hypothesis at a time.
  5. Apply the smallest fix that addresses the verified cause.
  6. Re-run the reproduction and any nearby regression tests.

Root cause, not symptom

Every step from symptom to cause should trace to runtime evidence (a log line, a value, a failing assertion) — not a guess. When a hypothesis is refuted, revert the probe and try the next one; don't stack guards.

  • Resist the masking guard. A nil-check or try/except that silences a crash without explaining why it was nil is a hidden root cause, not a fix.
  • Restart bugs: suspect state before code. If a bug only appears after a restart, suspect state, not logic — code doesn't change between runs, state does. If clearing a state file restores behavior, the bug is in how that state is written or validated.
  • Cheap test target? Hand off to trove-tdd to pin the bug with a failing-test-first cadence before fixing.

Decision Gate: reproduction

Context: A fix without reproduction can mask the bug or create a second one. Question: Do we have a concrete reproduction? Options:

  • A. Yes - proceed to narrow and fix.
  • B. No - create or request a reproduction first.
  • C. The issue is intermittent - collect logs or add targeted instrumentation. Default: B, because verified bugs beat guessed fixes.

Output

Report the reproduction, root cause, fix, and verification. If verification is partial, say exactly what is still unproven.

04 同仓库更多 Skills