* migrate .spell DSL from indentation to brace-delimited syntax (SPEC-001)
Replace Python-style 2-space indentation with Rust/Go curly-brace
blocks across the entire codebase. This is a breaking syntax change
that improves agent authorship — LLMs handle braces far better than
whitespace-significant formats.
Tokenizer: remove INDENT/DEDENT tokens, split bracketDepth into
parenDepth (suppresses newlines) and braceDepth (preserves newlines),
remove IndentationError.
Parser: replace all INDENT/DEDENT patterns with LBRACE/RBRACE across
~22 methods. New parseBraceBlock() helper replaces parseStatementBlock().
Two patterns: label blocks (label: { }) and keyword blocks (keyword { }).
AST, transformer, IR generator, and runtime are unchanged — indentation
was purely a lexical/parse concern.
Updated 86 .spell fixture files, 15 test files, 20 documentation files,
README, AGENTS.md, VM spec references, and compiler-pipeline docs.
620 tests pass, 0 fail.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Add changeset
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* Add impl plan for defihack
* add Yellow + LI.FI multi-track support with ENS hydration and e2e demo runbook
- implement end-to-end custom action support in compiler, IR generation, runtime action resolution, and executor routing
- add new offchain venue adapters: and , and register/export them in
- add adapter test suites for Yellow session lifecycle and LI.FI swap/bridge/compose_execute flows
- add LI.FI compose_execute safety guard: enforce unless explicit override is set
- add CLI ENS profile hydration (, ) for and
- clamp ENS-hydrated to safe bounds (0..500) in
- remove unused/dead ENS helper surface
- add DefiHack demo spells and prompt-first README runbook (VM rehearsal + deterministic CLI e2e)
- update README and reference docs for new venues and CLI flags
* Add changeset
* Remove unused docs
Fran Algaba