SKILL.md
Rigorous Validation
Intent
- Ship only when behavior is proven, not assumed.
- Treat tests, QA scripts, linters, and on-chain/off-chain simulations as first-class deliverables.
- Capture evidence so reviewers can verify quickly.
Inputs
- Canonical test commands (unit, integration, e2e, contract sims, UI snapshots).
- Acceptance criteria plus measurable signals (logs, screenshots, transaction hashes).
- Migrations/seed data steps required to exercise the change locally.
Workflow
- Design tests before coding
- Specify failing cases and target assertions for each requirement. - Align on how to stub external services, wallets, or platform APIs.
- Automate and isolate
- Prefer deterministic, headless test harnesses; avoid manual-only steps. - Seed data/fixtures close to tests to prevent global coupling.
- Run the full relevant matrix
- Cover affected platforms (iOS/Android/web), runtimes, or chain environments. - Capture artifacts (logs, screenshots, traces) for any non-deterministic checks.
- Track outcomes
- Record exact commands run and their commit hash in the PR or issue. - File follow-ups for flaky tests before merging.
Verification
- All targeted tests green and documented; no TODOs or skipped suites without owner sign-off.
- Static analysis, type-checking, formatters, and contract analyzers pass.
- Evidence (artifacts, hashes, screenshots) attached or linked for reviewer inspection.