reality-verification

What problem does it solve?

Reproducing intermittent or persistent failures and proving that a code change actually resolves the original problem can be slow, undocumented, and error prone. This Skill provides a repeatable BEFORE/AFTER verification process so teams can reproduce failures, run targeted reproduction commands, and record structured verification outcomes to avoid regressions and false positives.

Core Features & Use Cases

• Failure reproduction mapping: Map goal keywords to appropriate reproduction commands such as gh run view --log-failed for CI, project test command for unit tests, pnpm check-types or tsc --noEmit for typing issues, pnpm lint for lint problems, and MCP tools for E2E or API checks.
• Before/After documentation: Standardize entries in .progress.md under "Reality Check (BEFORE)" and "Reality Check (AFTER)" to capture reproduction commands, observed outputs, timestamps, and verification verdicts.
• Test quality validation: Run heuristic checks for mock-heavy or fragile tests and flag mock-quality anti-patterns before marking a fix as verified.
• Use case example: For a failing auth test, record the failing output and timestamp, rerun tests after the change, compare outputs, perform mock-quality checks, and document verification success or required follow-ups.

Quick Start

Ask the skill to reproduce and verify a reported failing test by providing the spec name, reproduction steps, and target command so it documents BEFORE and AFTER verification.