ad-spike

Plan a spike for an unknown technique by identifying uncertain scenarios and drafting a staged evaluation plan. This approach aligns with WORKFLOW.md §14, using golden fixtures to ensure repeatable comparisons and traceable outcomes at each checkpoint.

Golden fixtures in spike workflows are stage-by-stage documented contracts that enable repeatable comparisons. They provide a stable baseline for validating candidate techniques, allowing teams to reveal divergence early and make auditable decisions based on consistent test data.

Use staged spikes when facing unknown-technique scenarios that require disciplined exploration before commitment. If a technique carries significant uncertainty, a gate-driven pipeline with per-stage checkpoints reveals divergence early, preventing costly implementation mistakes down the line.

Document sources during a technical spike by citing official docs and public references with repository paths, URLs, or notes. Survey in-repo analogs to ground the exploration, ensuring the evaluation plan references existing codebase patterns and verified external documentation.

The best way to evaluate a spike outcome is using a two-layer evaluation combining end-to-end validation and per-stage analysis. This gate-driven approach produces an ADR to formally promote or delete the spike based on auditable, milestone-ready artifacts.

Staged spikes are designed to align with WORKFLOW.md §14, so having this file provides the canonical structure for discovery and gate-driven evaluation. Without it, you would need to establish equivalent milestone checkpoints and fixture lifecycle rules manually.