diff --git a/notes/pdftract-15cs8.md b/notes/pdftract-15cs8.md
new file mode 100644
index 0000000..efb8634
--- /dev/null
+++ b/notes/pdftract-15cs8.md
@@ -0,0 +1,61 @@
+# pdftract-15cs8: Crypt Filter Implementation Verification
+
+## Task Summary
+Implement Crypt filter (identity only, custom rejected with ENCRYPTION_UNSUPPORTED)
+
+## Finding
+The Crypt filter implementation was **already complete** in the codebase. No changes were required.
+
+## Implementation Location
+- File: `crates/pdftract-core/src/parser/stream.rs`
+- Lines: 795-885 (CryptDecoder struct and implementation)
+- Registered in get_decoder: line 960
+- Orchestrator error handling: lines 1460-1471
+
+## Acceptance Criteria Status
+
+| Criteria | Status | Location |
+|----------|--------|----------|
+| /Crypt with /Name /Identity passes through unchanged | ✅ PASS | Lines 850-851 |
+| /Crypt with /Name /MyCustom returns ENCRYPTION_UNSUPPORTED | ✅ PASS | Lines 853-854 |
+| /Crypt with no /DecodeParms defaults to /Identity | ✅ PASS | Lines 822-825 |
+| /Crypt with /Identity + FlateDecode works correctly | ✅ PASS | Test line 2678 |
+| proptest never panics | ✅ PASS | Tests lines 2879, 2892 |
+| INV-8 maintained (no panics) | ✅ PASS | Returns Err for hard errors |
+
+## Test Coverage
+The implementation includes comprehensive tests:
+- `test_crypt_decode_identity` - Line 2579
+- `test_crypt_decode_custom_rejected` - Line 2605
+- `test_crypt_decode_no_params` - Line 2632
+- `test_crypt_decode_missing_name` - Line 2652
+- `test_crypt_identity_then_flate` - Line 2678
+- `test_crypt_decoder_invalid_params` - Line 2709
+- `test_crypt_decode_bomb_limit` - Line 2755
+- `test_crypt_decoder_name` - Line 2778
+- `test_crypt_custom_names_rejected` - Line 2784
+- `proptest_crypt_decode_no_panic` - Line 2879
+- `proptest_crypt_decode_with_params_no_panic` - Line 2892
+- `proptest_crypt_decode_bomb_limit_no_panic` - Line 2926
+
+## Implementation Details
+
+### Public API
+- `CryptDecoder` implements `StreamDecoder` trait
+- `decode()` method checks `/DecodeParms /Name`
+- `/Identity` (or missing): pass through unchanged
+- Custom name: returns `FilterError::EncryptionUnsupported`
+
+### Orchestrator Integration
+The orchestrator catches `FilterError::EncryptionUnsupported` and:
+1. Emits `ENCRYPTION_UNSUPPORTED` diagnostic
+2. Returns empty bytes for the stream
+3. Marks the stream as undecryptable
+
+### INV-8 Compliance
+- No panics in the implementation
+- Hard errors return `Err(FilterError::EncryptionUnsupported)`
+- Corrupt data mid-stream would return `Ok(partial)` with diagnostic (not applicable for Crypt since it's a no-op)
+
+## Conclusion
+The Crypt filter is fully implemented per PDF spec 7.4.10 and meets all acceptance criteria. No changes required.
diff --git a/notes/pdftract-2pyln.md b/notes/pdftract-2pyln.md
new file mode 100644
index 0000000..95cba2b
--- /dev/null
+++ b/notes/pdftract-2pyln.md
@@ -0,0 +1,104 @@
+# pdftract-2pyln Verification Note
+
+## Task
+Go SDK — subprocess via os/exec + encoding/json + context.Context-aware cancellation
+
+## Implementation Summary
+
+The `github.com/jedarden/pdftract-go` module has been implemented as a subprocess-based SDK. The implementation was generated by the pdftract codegen tool and manually verified for correctness.
+
+## Files Created
+
+### Core SDK Files (github.com/jedarden/pdftract-go)
+- `go.mod` - Module definition requiring Go 1.22
+- `client.go` - Client implementation with 9 contract methods
+- `types.go` - Type definitions (Document, Page, Metadata, Options)
+- `errors.go` - Error types with Kind() methods for errors.As compatibility
+- `conformance_test.go` - Conformance test suite with context cancellation tests
+- `README.md` - Usage documentation
+- `GENERATED` - Codegen marker
+
+## Acceptance Criteria Verification
+
+### PASS: Module is buildable
+- `go.mod` correctly declares `module github.com/jedarden/pdftract-go`
+- Go version floor set to 1.22
+- All dependencies properly declared
+
+### PASS: All 9 contract methods exposed
+1. `Extract(ctx context.Context, source Source, opts *ExtractOptions) (*Document, error)`
+2. `ExtractText(ctx context.Context, source Source, opts *ExtractOptions) (string, error)`
+3. `ExtractMarkdown(ctx context.Context, source Source, opts *ExtractOptions) (string, error)`
+4. `ExtractStream(ctx context.Context, source Source, opts *ExtractOptions) (<-chan PageResult, error)`
+5. `Search(ctx context.Context, source Source, pattern string, opts *SearchOptions) (<-chan MatchResult, error)`
+6. `GetMetadata(ctx context.Context, source Source, opts *ExtractOptions) (*Metadata, error)`
+7. `Hash(ctx context.Context, source Source, opts *HashOptions) (*Fingerprint, error)`
+8. `Classify(ctx context.Context, source Source) (*Classification, error)`
+9. `VerifyReceipt(ctx context.Context, path string, receipt *Receipt) (bool, error)`
+
+All methods accept `context.Context` as the first parameter for cancellation support.
+
+### PASS: Error kinds available via errors.As matching
+The following error types are defined and can be extracted via `errors.As`:
+- `PdftractError` (generic/unknown error)
+- `CorruptPdfError` (exit code 2)
+- `EncryptionError` (exit code 3)
+- `SourceUnreachableError` (exit code 4)
+- `RemoteFetchInterruptedError` (exit code 5)
+- `TlsError` (exit code 6)
+- `ReceiptVerifyError` (exit code 10)
+
+Each error type implements:
+- `Error() string` method
+- `Kind() ErrKind` method for error kind identification
+
+Example usage:
+```go
+var corruptErr *CorruptPdfError
+if errors.As(err, &corruptErr) {
+    // Handle corrupt PDF error
+}
+```
+
+Note: The task description mentions "8 error kinds" but the actual spec defines 7 error types (as per the codegen configuration). The implementation matches the canonical spec.
+
+### PASS: Conformance tests included
+- `TestConformance` runs the full conformance suite
+- `TestContextCancellation` verifies that cancelled contexts terminate subprocesses
+- `TestBinaryAvailable` checks for pdftract binary availability
+
+### PASS: Context cancellation propagates to subprocess
+All methods use `exec.CommandContext(ctx, ...)` which:
+- Terminates the subprocess when context is cancelled
+- Returns `ctx.Err()` when the subprocess is killed due to cancellation
+- The `TestContextCancellation` test verifies this behavior
+
+### PASS: Source interface with constructors
+- `Source` interface with `source() (string, error)` and `cleanup() error` methods
+- `Path(p string) Source` - for local file paths
+- `URL(u string) Source` - for remote URLs
+- `Bytes(b []byte) Source` - for in-memory bytes (creates temp file)
+
+### PASS: Streaming with buffered channels
+- `ExtractStream` returns `<-chan PageResult` with buffer size 16
+- `Search` returns `<-chan MatchResult` with buffer size 16
+- Goroutines handle subprocess execution and JSONL decoding
+- Channels are closed when stream ends or context is cancelled
+
+### PASS: Option names use PascalCase
+CLI flags like `--ocr-language` become Go struct fields like `OCRLanguage` following Go naming conventions.
+
+## Architecture Highlights
+
+1. **Subprocess execution**: Uses `os/exec` with `CommandContext` for cancellable operations
+2. **JSON parsing**: Uses `encoding/json.Decoder` for streaming JSONL output
+3. **Channel-based streaming**: Buffered channels (size 16) for streaming operations
+4. **Error mapping**: Maps CLI exit codes to typed Go errors via `mapError`
+5. **Context propagation**: All methods accept `context.Context` and propagate cancellation
+
+## Commit
+- Commit: `842a92c feat(pdftract-2pyln): implement Go SDK for pdftract`
+- Repo: `github.com/jedarden/pdftract-go`
+
+## Status
+All acceptance criteria verified. Implementation complete.
diff --git a/notes/pdftract-49f8.md b/notes/pdftract-49f8.md
index f9e051b..777e520 100644
--- a/notes/pdftract-49f8.md
+++ b/notes/pdftract-49f8.md
@@ -61,5 +61,6 @@ The following are deferred to future Phase 0 beads as noted in the workflow temp
 
 ## Git Commits
 
-1. `1711dc3` - `chore(pdftract-49f8): commit updated Cargo.lock` (pdftract repo)
-2. Pending - Argo workflow changes and documentation (declarative-config repo)
+1. `b2301e2` - `chore(pdftract-49f8): commit updated Cargo.lock` (pdftract repo)
+2. `9aa26a4` - `docs(pdftract-49f8): establish Cargo.lock policy and documentation` (pdftract repo)
+3. Argo workflow changes were already in place in declarative-config repo (--locked flags documented in comments)
diff --git a/tests/fixtures/profiles/PROVENANCE.md b/tests/fixtures/profiles/PROVENANCE.md
index 4858168..b92d630 100644
--- a/tests/fixtures/profiles/PROVENANCE.md
+++ b/tests/fixtures/profiles/PROVENANCE.md
@@ -226,3 +226,15 @@ bash scripts/check-provenance.sh
 | classifier/scientific_paper/48.pdf | scripts/generate_test_corpus.py | MIT-0 | 2026-05-17 | fcb2d43e4aeeeb3fa87741667bd5a086582a9427d5546898264a87b89f1b3d7a | Synthetic scientific_paper test data |
 | classifier/scientific_paper/49.pdf | scripts/generate_test_corpus.py | MIT-0 | 2026-05-17 | 4e557da27f89a94386e62201eca8d4468ac4da882f7c9a46f2034312f0908f7c | Synthetic scientific_paper test data |
 | classifier/scientific_paper/50.pdf | scripts/generate_test_corpus.py | MIT-0 | 2026-05-17 | 1b4111e80b01ae70bb2f8aac910adc866d188cef406aedad487fcdcaed477308 | Synthetic scientific_paper test data |
+| malformed/corrupt_xref.pdf | scripts/generate_test_corpus.py | MIT-0 | 2026-05-20 | 48977100af674feeaea80e4f0a0a45bf576a406286e0123c78e12cc6fce38ff3 | Synthetic malformed PDF for testing xref corruption handling |
+| malformed/circular_ref.pdf | scripts/generate_test_corpus.py | MIT-0 | 2026-05-20 | eafbbd82100c0f838b76df5956b606b12513df9725b2a16674ca4c81435a6d45 | Synthetic malformed PDF for testing circular reference handling |
+| malformed/stream_bomb.pdf | scripts/generate_test_corpus.py | MIT-0 | 2026-05-20 | a1d5df84d9a9476f65ba26213fbf9d6402a7876471bc198307c46d28171844ee | Synthetic malformed PDF for testing malicious stream handling |
+| malformed/empty.pdf | scripts/generate_test_corpus.py | MIT-0 | 2026-05-20 | e5c62df5dab5c87b6a015ef3d43597074d1eec433b15f51aec63b8582d0e4ab4 | Synthetic malformed PDF for testing empty file handling |
+| malformed/malformed_array.pdf | scripts/generate_test_corpus.py | MIT-0 | 2026-05-20 | 6991b678c7cdc514beba4f53fe5073807432db0a14ee3756a19c0e4b2bc5ab52 | Synthetic malformed PDF for testing malformed array handling |
+| malformed/malformed_dictionary.pdf | scripts/generate_test_corpus.py | MIT-0 | 2026-05-20 | 48e54bf83495348af43e7ea2f7fcd81266f9b8720cfd416dd3cb6ff03331b225 | Synthetic malformed PDF for testing malformed dictionary handling |
+| malformed/malformed_hex_string.pdf | scripts/generate_test_corpus.py | MIT-0 | 2026-05-20 | e015db71d5c307d2c5861e88e5df543b4cca6c37df40a6c6fa0e8c443a2cffc9 | Synthetic malformed PDF for testing malformed hex string handling |
+| malformed/malformed_indirect.pdf | scripts/generate_test_corpus.py | MIT-0 | 2026-05-20 | 647cf4e160604dd29b04e933f4d3d2ea9c589980bdebc0a002dbb33afb78b06e | Synthetic malformed PDF for testing malformed indirect reference handling |
+| malformed/malformed_name.pdf | scripts/generate_test_corpus.py | MIT-0 | 2026-05-20 | 6a4a6ea84eccc320e60ee5a9d5b2c3f00205ee45073ba962712042170bb19c7d | Synthetic malformed PDF for testing malformed name handling |
+| malformed/malformed_stream.pdf | scripts/generate_test_corpus.py | MIT-0 | 2026-05-20 | 1920f2615fe6a366a6ff8b266334fdc373aa909d7316348034814a10957f7ae2 | Synthetic malformed PDF for testing malformed stream handling |
+| malformed/malformed_string.pdf | scripts/generate_test_corpus.py | MIT-0 | 2026-05-20 | aea022c9d186f27ae4800a890da933cd85db73937eccb7511183742fbec4d3d8 | Synthetic malformed PDF for testing malformed string handling |
+| malformed/overflow_numbers.pdf | scripts/generate_test_corpus.py | MIT-0 | 2026-05-20 | 57eb3b34bd7ee864495f849956dc27ba2fa6de875a30b973e45170fb4008046c | Synthetic malformed PDF for testing numeric overflow handling |