chore: sync changes from prior session

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: hyperpolymath

PROJECT_MANAGEMENT/CURRENT_STATE_INVENTORY.md

@@ -0,0 +1,25 @@
+# PROOF-NEEDS.md — palimpsest-license
+
+## Current State
+
+- **src/abi/*.idr**: NO
+- **Dangerous patterns**: 1 `unwrap()` in Rust tools
+- **LOC**: ~11,800 (Rust tools + license text)
+- **ABI layer**: Missing
+
+## What Needs Proving
+
+| Component | What | Why |
+|-----------|------|-----|
+| pmpl-verify | Signature verification is correct (no false positives or negatives) | False positive verification accepts tampered licenses |
+| pmpl-sign | Signing produces valid, verifiable signatures | Invalid signatures break downstream verification |
+| pmpl-audit | Audit correctly identifies all license violations | Missed violations create legal exposure |
+| License text parsing | Parser correctly extracts SPDX identifiers | Wrong SPDX extraction causes compliance errors |
+
+## Recommended Prover
+
+**Idris2** — Create `src/abi/` with types for signature verification correctness. The sign/verify pair should be proven as inverses.
+
+## Priority
+
+**LOW** — License tooling. While correctness matters, the blast radius of bugs is limited to incorrect license reports rather than security or data integrity failures.

PROOF-NEEDS.md

@@ -0,0 +1,57 @@
+# TEST-NEEDS.md — palimpsest-license
+
+> Generated 2026-03-29 by punishing audit.
+
+## Current State
+
+| Category     | Count | Notes |
+|-------------|-------|-------|
+| Unit tests   | 6     | Haskell Spec: BilingualSpec, LicenseSpec, MetadataSpec, ReferenceSpec, UtilsSpec + 1 Integration/PipelineSpec |
+| Integration  | 1     | Integration/PipelineSpec.hs |
+| E2E          | 0     | None |
+| Benchmarks   | 0     | None |
+
+**Source modules:** ~18 across multiple languages. Haskell validator: 8 modules (Main, Validator, Bilingual, License, Metadata, Reference, Types, Utils). Elixir: 5 Ecto modules (derivative, license, lineage, violation, work) in ARCHIVE. OCaml: 1 test. Rust: 6 files. PHP: 4 files.
+
+## What's Missing
+
+### P2P (Property-Based) Tests
+- [ ] License validator: property tests for arbitrary license text classification
+- [ ] Bilingual validator: property tests for language detection accuracy
+- [ ] Metadata validator: arbitrary SPDX header validation
+- [ ] Reference validator: property tests for citation format compliance
+
+### E2E Tests
+- [ ] Full validation pipeline: input file -> detect license -> validate -> report
+- [ ] Multi-file: validate entire repository license compliance
+- [ ] Cross-language: Haskell validator agrees with Rust validator agrees with OCaml validator
+
+### Aspect Tests
+- **Security:** No tests for malicious license text injection, path traversal in file scanning
+- **Performance:** No validation throughput benchmarks
+- **Concurrency:** No tests for parallel file scanning
+- **Error handling:** No tests for binary files, empty files, files with mixed encodings
+
+### Build & Execution
+- [ ] `cabal test` or `stack test` for Haskell
+- [ ] `cargo test` for Rust
+- [ ] OCaml test execution
+
+### Benchmarks Needed
+- [ ] License detection speed per file
+- [ ] Batch validation throughput (files/second)
+- [ ] Bilingual detection accuracy and speed
+
+### Self-Tests
+- [ ] Validate its own LICENSE file
+- [ ] Validate its own SPDX headers across all source files
+
+## Priority
+
+**MEDIUM.** The Haskell validator has 7 test files for 8 modules — actually decent coverage ratio. But the Rust, OCaml, PHP, and Elixir implementations have ZERO tests. The cross-language consistency is completely unverified. No benchmarks for a tool that needs to scan entire repositories quickly.
+
+## FAKE-FUZZ ALERT
+
+- `tests/fuzz/placeholder.txt` is a scorecard placeholder inherited from rsr-template-repo — it does NOT provide real fuzz testing
+- Replace with an actual fuzz harness (see rsr-template-repo/tests/fuzz/README.adoc) or remove the file
+- Priority: P2 — creates false impression of fuzz coverage