docs: s-expression artifact format — honest cost analysis, recommend Option A

[avrabe]

What this ships

`docs/design/sexpr-artifact-format.md` — honest cost analysis for adding s-expression as an artifact format alongside YAML.

TL;DR

Recommended path: Option A (`rivet convert --to sexpr`, one-way export-only), gated on the YAML fuzzer (#158 in this batch).

• Option A (~5 days): one-way export. Audit-archive use. No LSP work. Ships regardless of fuzzer outcome.
• Option B (~15-20 days): read-only import. Only if fuzzer finds real bugs on our corpus.
• Option C (40-60 days + ~20%/yr dual-path maintenance): not justified on current evidence.
• Option D (replace YAML): not recommended.

Total for recommended: ~5 (Option A) + ~2 (fuzzer) ≈ 7 days.

Why not further

1. Paper honesty: arxiv:2604.13108 shows LLM comprehension is format-agnostic. JSON has lower silent-corruption (21%) than both YAML and S-expr (50% each). The paper's own proposal is a single top-level `intent.lisp`, not per-artifact migration.
2. What we already have (verified by reading code, not guessing):
   • `sexpr.rs` (579 LoC) is rowan-based and lossless — comment/whitespace preservation free
   • `yaml_hir.rs::extract_schema_driven` is 1901 LoC — its s-expr twin is the single largest item (~900-1200 LoC)
   • `yaml_edit.rs` is 1210 LoC line-based — s-expr twin is cleaner but still ~800-1000 LoC
   • 7 call sites filter on `.yaml`/`.yml` extensions (enumerated)
   • vscode-rivet has no TextMate grammar — `.rivet` grammar is new work
3. 3 of 8 Mythos findings retired by s-expr ingest (null shorthand, multi-doc truncation, Norway problem). The other 5 are orthogonal.

Test plan

Doc only.

🤖 Generated with Claude Code

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Rivet Criterion Benchmarks'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: aef153e	Previous: aa706fc	Ratio
link_graph_build/10000	37599747 ns/iter (± 2642609)	27530692 ns/iter (± 3638629)	1.37
validate/10000	15082659 ns/iter (± 1425852)	10896266 ns/iter (± 1631891)	1.38
query/10000	146593 ns/iter (± 676)	90574 ns/iter (± 529)	1.62

This comment was automatically generated by workflow using github-action-benchmark.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!