maude: Phase 6 strategy language (19 tests, 178 total)

lib/maude/strategy.sx — first-class set-valued strategies: idle/fail/all/
rule/seq/alt/star/plus/bang/name combinators, named-strategy env. Same
rule set computes different things under different strategies; verified
with single-rule vs all vs seq-order vs alt vs star vs bang.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

plans/maude-on-sx.md

@@ -92,9 +92,9 @@ The novel substrate stress: equational matching. Pattern `X + Y` against `1 + 2
 - [x] Tests: state-transition systems (puzzle solvers, protocol simulators).
 
 ### Phase 6 — Strategy language
-- [ ] Compose strategies: sequential `;`, alternative `|`, iteration `*`, fixed-point.
-- [ ] User-named strategies; strategy expressions as values.
-- [ ] Tests: programs whose meaning depends on strategy choice.
+- [x] Compose strategies: sequential `;`, alternative `|`, iteration `*`, fixed-point.
+- [x] User-named strategies; strategy expressions as values.
+- [x] Tests: programs whose meaning depends on strategy choice.
 
 ### Phase 7 — Reflection (META-LEVEL)
 - [ ] Terms-as-data: `META-LEVEL` lets you encode/decode terms as Maude terms.
@@ -220,5 +220,20 @@ The novel substrate stress: equational matching. Pattern `X + Y` against `1 + 2
     `.` can't be an op token (statement terminator). `mau/search` is the prime
     Phase 7 reflection / Phase 8 extraction target alongside the matcher.
 
+- **Phase 6 (strategy language) — DONE, 178/178 total.**
+  `lib/maude/strategy.sx`. Strategies are first-class tagged-dict VALUES and
+  set-valued: `mau/sapply ctx strat term` → deduped (by canon) list of results.
+  Combinators: `idle`/`fail`/`all`/`rule LABEL`/`seq`/`alt`/`star`/`plus`/`bang`
+  /`name`. `seq` = flatmap B over A's results; `alt` = union; `star` = reflexive-
+  transitive closure (BFS, canon-deduped); `plus` = A then star; `bang` =
+  normal forms (reachable terms where A yields nothing). Named strategies via a
+  NAME->strategy env dict passed to `mau/srun`/`mau/srun-canon`. Verified that
+  the same rule set computes different things under different strategies
+  (single rule vs all vs seq order vs alt vs star vs bang). Built on Phase 5
+  `mau/all-successors` (rule label filter = `mau/rules-with-label`).
+  - Note: `dict-set!` returns the value, not the dict — build a named-strategy
+    env by binding `(define env {})` then `(dict-set! env ...)`, pass `env`.
+    `srun-canon` sorts results so expected lists must be sorted.
+
 ## Blockers
 _(none)_