Answers the one question before you ship: May I ship? Catch packaging traps, dependency conflicts, and import leaks before CI goes green and a broken wheel hits production.
Tip: Use ←/→ arrow keys or mouse wheel to navigate. Tip: swipe horizontally to browse.
AI assistants make a repo look finished overnight. Tested blocks vanish, the same logic appears many times, and past a certain file size the model slows down while analysis cost climbs and quality drops. ManifestGuard reports only: complexity budgets and breaks of defined refactoring rules (not a line-count cutoff). Split and correction are for a human or, best case, the AI.
Checks pyproject.toml, setup.py, and requirements.txt for consistency. Finds typical packaging traps before the wheel is built.
Finds broken script targets, undeclared third-party imports, and version conflicts before a user hits a crash.
Apply package allow/block rules and version constraints from plain configuration.
The free check finds unused, unordered, undeclared, and circular imports before they disturb a release. Commands are in the guides.
Unlock complexity, duplication, dummy/placeholder code, and related health metrics.
Keeps CI from going green on stale coverage. Deterministic exit codes for the pipeline. Setup is in the guides.
ManifestGuard Pro, second test branch — not inside the everyday check: forgotten thread, timer, logger, and QThread patterns in source. Command: check-lifecycle. How to run it is in the guides.
Machine-readable reports (SARIF/JSON) and SBOM completeness for audits. SPDX and CycloneDX are standard formats for these inventories.
Preview supported fixes first. ManifestGuard writes only registered patches after a workspace snapshot and file backups, where the license allows it.
Classic linters (like ruff) edit your code. Security scanners look for known CVEs. ManifestGuard Python is the quality gate before you ship.
When pyproject.toml does not match the real import tree, script targets go nowhere, or coverage numbers in CI are stale, that usually shows up after release. The everyday ManifestGuard check is for that: packaging, entry points, dependencies, and import hygiene — the quality gate before you ship.
ManifestGuard Pro goes two steps further, and they stay separate. --extended deepens the same check (complexity, honest CI coverage, SBOM). check-lifecycle is a second test branch: leak patterns in the source (threads, timers, loggers). That leak scan is not inside the regular check and not inside --extended.
AI assistants are fast and error-prone. Tested blocks vanish, redundancies multiply, and past a certain file size the model slows down. Bloated code raises analysis cost while quality drops and debugging takes longer. ManifestGuard reports complexity budgets and breaks of defined refactoring rules (not a line-count cutoff). Split and correction are done by a human or, best case, by the AI. Then run the check again until the ship gate is clean.
ManifestGuard Python is the AI dream: find weaknesses in development without GitHub Actions or other expensive test tools, and have them corrected. ManifestGuard reports only. Split and correction follow defined refactoring rules — by a human or, best case, by the AI.
The Community version is the everyday check. ManifestGuard Pro unlocks the deeper analysis, the separate leak scan, machine-readable SBOM/evidence reports, and 2 device activations (laptop + CI). The license is annual with 12 months of updates; after that you can keep using the last entitled version.
Install, license activation, and every command live in the guides.
|
Manifest & packaging validation
|
✓
|
✓
|
|
Entry point consistency checks
|
✓
|
✓
|
|
Dependency analysis
|
✓
|
✓
|
|
Import hygiene
|
✓
|
✓
|
|
Custom validation rules
|
–
|
✓
|
|
Extended analysis
|
–
|
✓
|
|
Coverage for CI
|
–
|
✓
|
|
Lifecycle leak scan
|
–
|
✓
|
|
SBOM & CVE signals
|
–
|
✓
|
|
QuickFix suggestions
|
–
|
✓
|
|
QuickFix apply (supported findings + backup)
|
–
|
–
|
Applies to the purchased version incl. all updates for 1 year from the purchase date.