# Project State

## Project Reference

See: .planning/PROJECT.md (updated 2026-02-25)

**Core value:** Archive impossible to unpack without knowing the format -- standard tools (7z, tar, unzip, binwalk) cannot recognize or extract contents
**Current focus:** Milestone v1.1 Directory Support

## Current Position

Phase: Not started (defining requirements)
Plan: —
Status: Defining requirements
Last activity: 2026-02-25 — Milestone v1.1 started

## Performance Metrics

**Velocity (from v1.0):**
- Total plans completed: 10
- Average duration: 3.6 min
- Total execution time: 0.6 hours

## Accumulated Context

### Decisions

Decisions are logged in PROJECT.md Key Decisions table.
Recent decisions affecting current work:

- Roadmap: Format spec must precede all implementation (all three decoders build against same spec)
- Roadmap: Obfuscation (XOR headers, encrypted TOC, decoy padding) deferred to Phase 6 after all decoders work without it
- Phase 1: IV stored only in TOC, not duplicated in data blocks (simplifies shell dd extraction)
- Phase 1: Same 32-byte key for AES-256-CBC and HMAC-SHA-256 in v1 (v2 will use HKDF)
- Phase 1: Magic bytes 0x00 0xEA 0x72 0x63 (leading null signals binary)
- Phase 1: HMAC scope = IV (16 bytes) || ciphertext (encrypted_size bytes)
- Phase 2: Used rand::Fill::fill() for IV generation (correct rand 0.9 API)
- Phase 2: Manual binary serialization with to_le_bytes/from_le_bytes (no serde/bincode)
- Phase 2: Filename-only entry names (not full paths) for archive portability
- Phase 2: HMAC failure skips file and continues; SHA-256 mismatch warns but writes
- Phase 2: Flags bit 0 set only when at least one file is actually compressed
- Phase 3: Library crate with pub mod re-exports for all 6 modules
- Phase 3: Unit tests embedded in modules via #[cfg(test)] (not separate files)
- Phase 3: hex-literal v1.1 for compile-time SHA-256 known-value assertions
- Phase 3: Corrected HMAC golden vector (openssl pipe+xxd produced wrong value; verified with file input and Python)
- Phase 3: cargo_bin! macro for non-deprecated assert_cmd binary resolution
- Phase 3: 11MB deterministic pseudo-random data for large file test (wrapping_mul Knuth hash)
- Phase 4: Single-file Kotlin decoder (ArchiveDecoder.kt) for simplicity and Android embeddability
- Phase 4: RandomAccessFile for seeking to data blocks instead of reading entire archive into memory
- Phase 4: HMAC failure skips file, SHA-256 mismatch warns but writes (matching Rust behavior)
- Phase 4: Kotlin signed byte handling with .toByte() for literals > 0x7F, contentEquals() for ByteArray comparison
- Phase 5: POSIX sh (not bash) for maximum busybox compatibility
- Phase 5: xxd/od auto-detection at startup for hex conversion
- Phase 5: Graceful HMAC degradation when openssl lacks -mac support
- Phase 5: Extract ciphertext to temp file before decryption (avoids pipe buffering issues)
- Phase 5: LC_ALL=C for predictable byte handling across locales
- Phase 5: All 6 cross-validation tests passed on first run -- decode.sh was correct as written
- Phase 5: Used sh (not bash) to invoke decode.sh in tests for POSIX compatibility validation
- Phase 6: Always enable all 3 obfuscation features (no --no-obfuscate flag in v1)
- Phase 6: Decoy padding range 64-4096 bytes per file (FORMAT.md allows up to 65535)
- Phase 6: Shared read_archive_metadata() helper for unpack/inspect de-obfuscation
- Phase 6: Two-pass TOC serialization for correct data_offsets with encrypted TOC size
- Phase 6: XOR bootstrapping in Kotlin uses and 0xFF masking on BOTH operands for signed byte safety
- Phase 6: Shell decoder writes de-XORed header to temp file, reuses existing read_hex/read_le_u16/read_le_u32
- Phase 6: Shell decoder TOC_FILE/TOC_BASE_OFFSET abstraction for encrypted vs plaintext TOC
- Phase 6: Shell decoder HMAC constructs IV from parsed hex via hex_to_bin (not archive position)

### Pending Todos

None yet.

### Blockers/Concerns

None.

## Session Continuity

Last session: 2026-02-25
Stopped at: Starting milestone v1.1 — defining requirements
Resume file: None