Issues

• fix(lsp): a hyphenated policy ID is indexed as its first token, so policies sharing a prefix collide

  ai-discoveredbug
  ymm-oss/fsl#1880 · Filter by author rizumita opened on Nov 13, 2026 · M25 層間 refine の保証強化

• refactor(cache): verify-cache format tags are spelled per reader/writer; chain restates the key rule in prose

  ai-discoveredenhancement
  ymm-oss/fsl#1877 · Filter by author rizumita opened on Nov 13, 2026 · M27 下準備: M25・M26 が触る重複の一本化

• refactor(cli): per-invocation state rides on a thread-local and two process-wide statics with different scopes

  ai-discoveredenhancement
  ymm-oss/fsl#1876 · Filter by author rizumita opened on Nov 13, 2026 · M27 下準備: M25・M26 が触る重複の一本化

• refactor(load): check and nine commands hand-write root admission; three loaders build the model differently

  ai-discoveredenhancement
  ymm-oss/fsl#1875 · Filter by author rizumita opened on Nov 13, 2026 · M27 下準備: M25・M26 が触る重複の一本化

• refactor(bmc): the per-instance definedness probe sequence is written in two loops with different filters

  ai-discoveredenhancement
  ymm-oss/fsl#1874 · Filter by author rizumita opened on Nov 13, 2026 · M27 下準備: M25・M26 が触る重複の一本化

• refactor(verifier): scoped SAT probes are hand-written per helper, and each folds solver Unknown its own way

  ai-discoveredenhancement
  ymm-oss/fsl#1873 · Filter by author rizumita opened on Nov 13, 2026 · M27 下準備: M25・M26 が触る重複の一本化

• refactor(test): each CLI test file spawns fslc with its own cache setup, and most inherit the user's verify cache

  ai-discoveredenhancement
  ymm-oss/fsl#1872 · Filter by author rizumita opened on Nov 13, 2026 · M27 下準備: M25・M26 が触る重複の一本化

• refactor(verify): scenarios, explain and mutate each restate which BmcResult is a failure

  ai-discoveredenhancement
  ymm-oss/fsl#1871 · Filter by author rizumita opened on Nov 13, 2026 · M27 下準備: M25・M26 が触る重複の一本化
  1

• refactor(refine): five fslc-rust/Worker consumers each map RefinementVerdict to a result word and exit

  ai-discoveredenhancement
  ymm-oss/fsl#1870 · Filter by author rizumita opened on Nov 13, 2026 · M27 下準備: M25・M26 が触る重複の一本化
  1

• perf(html,explain): agentic_rag_design takes 17 minutes in html at the default depth - 97 times verify

  ai-discoveredenhancement
  ymm-oss/fsl#1868 · Filter by author rizumita opened on Nov 12, 2026 · M26 速度: 再計算の重複と資源の上限

• refactor(cli): main.rs keeps copies of library functions (strict-tag warnings, violation bindings/blame JSON, log-record mapping), and each copy serves a different command

  ai-discoveredenhancement
  ymm-oss/fsl#1865 · Filter by author rizumita opened on Nov 12, 2026 · M11 内部品質: 単一オーナーシップと契約整合

• refactor(syntax,core): ParseOptions carries no data since #1434, yet 19 X / X_with_options pairs and 373 mentions keep it threaded through every entry

  ai-discoveredenhancement
  ymm-oss/fsl#1864 · Filter by author rizumita opened on Nov 12, 2026 · M22 コード衛生: Rust 慣行と静的検査

• fix(domain): domain generate refuses documents domain check verifies when an aggregate name has consecutive capitals or a state field is camelCase

  ai-discoveredbug
  ymm-oss/fsl#1863 · Filter by author rizumita opened on Nov 12, 2026 · M20 domain lowering の単一化
  1

• perf(analyze): action_dependency_graph builds pairwise edges per state variable - 7 s and 3 GB when 800 actions share one variable

  ai-discoveredenhancement
  ymm-oss/fsl#1858 · Filter by author rizumita opened on Nov 11, 2026 · M26 速度: 再計算の重複と資源の上限

• perf(core): a decision table whose domain is near the enumeration bound costs 1.2 s per lowering, and every command pays it one to three times (check 1.9 s, document generate 3.1 s, fmt --check 1.3 s)

  ai-discoveredenhancement
  ymm-oss/fsl#1857 · Filter by author rizumita opened on Nov 11, 2026 · M26 速度: 再計算の重複と資源の上限

• perf(core): nested def calls expand exponentially - check does not finish on 24 two-call definitions

  ai-discoveredbug
  ymm-oss/fsl#1856 · Filter by author rizumita opened on Nov 11, 2026 · M26 速度: 再計算の重複と資源の上限

• fix(core,verifier): a large numeric range aborts with an allocation failure or a stack overflow instead of a located error (verify, kernel, scenarios, check)

  ai-discoveredbug
  ymm-oss/fsl#1855 · Filter by author rizumita opened on Nov 11, 2026 · M26 速度: 再計算の重複と資源の上限

• perf(playground): each Verify creates a new Worker and re-initialises Z3 and the wasm module; the Worker verify re-parses the source and runs two concrete searches

  ai-discoveredenhancementneeds-evidence
  ymm-oss/fsl#1854 · Filter by author rizumita opened on Nov 11, 2026 · M26 速度: 再計算の重複と資源の上限

• perf(lsp): no index or diagnostics cache per document version - every request re-parses, semantic tokens re-resolve per reference, and the requirements solver runs on each edit

  ai-discoveredenhancementneeds-evidence
  ymm-oss/fsl#1853 · Filter by author rizumita opened on Nov 11, 2026 · M26 速度: 再計算の重複と資源の上限

• perf(frontend): one invocation parses, lowers and builds the model of the same source many times (check lowers twice and lexes the root about 16 times)

  ai-discoveredenhancementneeds-evidence
  ymm-oss/fsl#1852 · Filter by author rizumita opened on Nov 11, 2026 · M26 速度: 再計算の重複と資源の上限
  1

• perf(runtime): the explicit engine, the concrete probes and conformance did not get the #1247 improvements - guards evaluated twice, the state cloned per transition, visited sets of full state copies

  ai-discoveredenhancementneeds-evidence
  ymm-oss/fsl#1851 · Filter by author rizumita opened on Nov 11, 2026 · M26 速度: 再計算の重複と資源の上限

• perf(cache): the verify cache is consulted only after the whole front end, sweep restarts every cell from source, and refine / inline implements have no cache

  ai-discoveredenhancementneeds-evidence
  ymm-oss/fsl#1850 · Filter by author rizumita opened on Nov 11, 2026 · M26 速度: 再計算の重複と資源の上限
  1

• perf(mutate): fixed cost per mutant - two BMC runs, contracts re-resolved from the unmutated source, the mutant lowered two or three times - and an uncached baseline

  ai-discoveredenhancementneeds-evidence
  ymm-oss/fsl#1849 · Filter by author rizumita opened on Nov 11, 2026 · M26 速度: 再計算の重複と資源の上限

• perf(verify): the --lemma loop re-runs an identical BMC base per adopted lemma, induction solves the ranking obligations twice, and plain BMC always runs a concrete search first

  ai-discoveredenhancementneeds-evidence
  ymm-oss/fsl#1848 · Filter by author rizumita opened on Nov 11, 2026 · M26 速度: 再計算の重複と資源の上限

• perf(verify): BMC issues solver checks per action instance and for obligations that are trivially true; attribute verify time by obligation kind

  ai-discoveredenhancementneeds-evidence
  ymm-oss/fsl#1847 · Filter by author rizumita opened on Nov 11, 2026 · M26 速度: 再計算の重複と資源の上限
  1