Archive Integrity Batch Checker
Validate ZIP, CBZ, and EPUB archives in bulk without permanently extracting them. The command uses explicit UTF-8 JSON input and produces reviewable JSON or Markdown output.
Three-minute start
python -m pip install .
archive-integrity examples/sample.json
archive-integrity examples/sample.json --format json --output report.json
The example documents the input shape. Version 1.1 warns about empty containers, unsafe member paths, and members above a configurable max_compression_ratio, while retaining the exact member evidence. Existing report files are never overwritten. Source inputs are read-only except where the documented purpose explicitly creates a new output artifact.
Privacy and platforms
The tool runs locally and does not upload input or include telemetry. Python 3.10 or newer is supported on Windows, macOS, and Linux.
Interpretation boundary
The tool validates ZIP-family containers only. Warnings are review evidence rather than malware findings; a structurally valid archive can still contain unsafe, misleading, or semantically invalid content.
Development
python -m pip install -e ".[dev]"
ruff format --check .
ruff check .
mypy src
pytest
python -m build
The project is feature-complete for its documented v1 scope. Maintenance focuses on correctness, security, compatibility, and well-supported input improvements.
Part of the Logan Pendragon Forge open-source collection. Licensed under the MIT License.
Streaming verification budgets
Version 1.2 reads members in chunks with positive integer max_member_bytes
(default 268435456) and max_total_bytes (default 1073741824, per archive).
Limits count decompressed bytes, including directory entries. Reading stops after
at most one extra byte to detect exhaustion. Exact-limit archives can verify.
A status of incomplete, budget_exhausted, and verification_complete: false
identify budget exhaustion; these archives never count as valid.
Encrypted, unsupported, and corrupt members report errors. No member is extracted.
Budgets limit decompressed output reads, not ZIP metadata memory or CPU time.
Input and report guarantees (1.2.1)
Input must be a JSON object with a non-empty paths array of non-empty strings.
max_compression_ratio must be a finite positive JSON number; byte budgets must
be positive integers, excluding booleans. Relative archive paths are interpreted
against the command's working directory. A missing or inaccessible archive gets
its own error record and later paths still run. Source symbolic links are followed.
Exit 0 means a report was generated; inspect invalid_count, incomplete_count and
each archive status to decide whether the collection passed. Invalid input or
report I/O exits 2.
Reports use exclusive creation and refuse existing files or symbolic links,
including dangling links. A file created after validation is preserved. Keep
source archives stable during verification; this is not a filesystem snapshot.
This guard handles accidental collisions, not a hostile concurrent filesystem.
Both release packages install README, DEVELOPMENT, TESTING, CHANGELOG and the
sample beneath share/doc/archive-integrity-batch-checker in the installation
prefix. See TESTING.md for reproducible acceptance.
Optional exit policies (1.3.0)
By default, exit 0 still means the report was generated, even when archives need attention. For scripts and scheduled checks, select the conditions that should return exit 1 after the full report is successfully written:
archive-integrity input.json --format json --fail-on invalid
archive-integrity input.json --fail-on invalid --fail-on warning --output report.md
Repeat --fail-on to combine policies. invalid matches any unverified archive, including corrupt, missing, unsupported or budget-incomplete inputs (invalid_count > 0). incomplete matches budget-exhausted verification specifically (incomplete_count > 0). warning matches any reported review warning (warning_count > 0), including warnings on otherwise verified archives. Repeating a policy has no additional effect. These choices do not change archive interpretation, budgets, report schema or report contents. Warnings remain review evidence, not malware findings.
Input/argument and report-I/O failures still return exit 2 and take precedence. A policy failure returns 1 only after output succeeds; review the report to see the affected archives. Existing report files and symbolic links remain refused, and source files remain read-only. No new runtime dependencies or extraction are introduced.
