shadow-mirrorlisted

# Shadow Mirror A validation framework that casts **shadows** (instrumented traces) and holds up **mirrors** (assertions that reflect expected behavior). Systems are decomposed into an operation tree; assertions accumulate into provable hypotheses about correctness. Line coverage tells you what *ran*. Shadow Mirror tells you what's *proven* — and what to test next. ## The seven-phase loop (SM-0 .. SM-6) ``` SM-0 HYPOTHESIZE → state a falsifiable claim about a symptom SM-1 INSTRUMENT → choose probes that can answer the SM-0 question SM-2 ASSERT → generate predicates that accumulate toward proof SM-3 EXECUTE → cast the shadow: run, collect traces / metrics / outcomes SM-4 DOCUMENT → emit a durable, content-addressable evidence receipt SM-5 REVIEW → meta-validate: coverage, assertion quality, soundness SM-6 ITERATE → refine and loop, or declare done ``` The loop runs in two directions on one engine: - **Forward — test planning** (SM-0 → SM-1 → SM-2, generatively): given code or a diff, produce a *plan* — the gaps to cover and the assertion stubs to cover them. - **Backward — coverage mapping** (SM-3 → SM-5, analytically): given existing tests, produce a *map* — the operation tree scored per node, per level, surfacing the blind spots line coverage hides. > The canonical phase definitions (inputs, outputs, receipt mapping, the phase-count contract) live in `docs/phases.md` in the shadow-mirror repository. This skill is the operational com