Skip to content

macOS Inspector 2.2.0 - Saved PDF review and history coverage

Latest

Choose a tag to compare

@bacescuandrei bacescuandrei released this 25 Oct 08:58

macOS Inspector 2.2.0

Released October 25, 2026. This release completes saved PDF review in the document workspace and makes history-loading limits visible. Existing Mac audit, browser extension review, and static PDF inspection remain available.

Getting started

  1. Download macos-inspector-2.2.0-macos.zip and SHA256SUMS below.
  2. Verify the archive's SHA-256 against SHA256SUMS, then extract it.
  3. Double-click macOS Inspector.command and keep the launcher window open.
  4. Open PDF Inspector, expand Recent PDF inspections, and choose Review saved inspection beside a supported report.

Requires macOS and Python 3.10 or newer. Core operation needs no account, administrator privileges, mandatory Full Disk Access, or external service. Start the new package with your existing report output directory to review earlier supported reports. The launcher and installation guide explain output selection and first-launch behavior. The package is not Apple-signed or notarized; do not disable Gatekeeper globally.

Review saved PDF evidence without the original document

Review saved inspection loads the saved JSON evidence into the same PDF workspace used for a new inspection. The notice identifies the recorded analysis date and Inspector version. Actions, scripts, destinations, attachment references, metadata, and analysis limits remain available, with links to the original exports when present. Keyboard focus moves to the saved-evidence notice, and long fields wrap on narrow screens.

No PDF is uploaded, reopened, rendered, or reanalyzed. Loading does not authenticate the report, establish the document's current security status, or apply current parser rules to an older result. For example, an analysis from 2.0.0 retains that version and its interpretation limits; review of the report does not silently convert it into a 2.2.0 analysis. Inspect the original again when current checks are needed.

Use Clear to remove displayed evidence without deleting a saved report. Choosing another report, selecting a new file, or clearing the workspace invalidates pending loads and sharing previews. Late responses cannot restore a previously selected result, and mismatched response identities are rejected.

Prepare a sharing copy from a saved inspection

After loading saved evidence, press Preview sharing copy. Expand All fields in this sharing copy, review the complete reduced summary, then download HTML or JSON. The original document is not required.

The existing field allowlist omits detailed document evidence. Include document SHA-256 remains off by default and requires explicit selection. Original reports remain unchanged; downloads are generated in the browser, with no sharing copy saved on the server. A reduced copy is not anonymization, an authenticated original, a fresh analysis, or a safety verdict. Counts and optional fingerprints can still identify the document. The sharing summary's tool version identifies the software generating that copy, not necessarily the version that inspected the PDF.

For example, you can reopen last week's recorded inspection and prepare a reduced handoff without uploading the PDF again. Keep the detailed original report for investigation and read its original analysis date and limitations before drawing conclusions.

Understand history coverage

The history panel now separates loading coverage from PDF findings:

  • Usable records remain visible when other reports cannot be read or validated.
  • Selected but unavailable reports, rejected PDF-named entries, and older candidates not loaded have separate counts.
  • Incomplete or interrupted directory enumeration is explicit.
  • History selection details explains the selection method and limits.
  • Failed refreshes remove stale rows and coverage details. Missing or inconsistent metadata from older servers is labeled unavailable rather than interpreted as zero.

Enumeration examines at most 5,000 directory entries, including unrelated files, and selects at most 25 canonical regular report candidates. Ordering uses filesystem modification time among examined candidates, not analysis date. Copying or touching a report can change its position. An invalid recent candidate still occupies a selection slot; older files are not read to fill the gap. Incomplete enumeration cannot establish which reports are newest in the entire directory.

For example, a list can show one usable report while explaining that another selected report could not be loaded. That is a partial history view, not evidence that the omitted document is safe or malicious. An empty list does not prove that no saved reports exist. No report is deleted, repaired, or reanalyzed.

Read boundaries and privacy

Saved review requires authenticated local API access. Report-download cookies alone cannot authorize the review API. Identifiers must match the canonical inspection filename shape, and JSON reads are limited to 2 MiB. Unsupported schemas, malformed display fields, invalid Unicode, oversized files, and symlinked reports are rejected. Nonregular files, including FIFOs, are rejected before reading saved-review or sharing evidence.

History diagnostics use counts and fixed explanations, without copying rejected filenames, paths, or raw exception text. Supported history rows and original reports still contain recorded filenames, fingerprints, and potentially sensitive evidence. No external destination is contacted, no original PDF is retained, and no host scan or process action starts when reviewing PDF history.

These checks are not report authentication, an atomic filesystem snapshot, or protection against a compromised same-user or administrative account. The PDF parser's existing limits remain: no rendering, code execution, decryption, destination contact, or complete final-revision reachability analysis. Recorded indicators do not establish execution or compromise; missing indicators do not establish safety.

Validation

All 169 Python tests, 94 browser tests across Chromium and WebKit, and 26 synthetic trust/response scenarios passed locally. Coverage includes authenticated read-only retrieval, original metadata retention, real saved-report review after page reload, reduced sharing previews without another upload, delayed and mismatched responses, mixed valid/invalid history, enumeration and selection limits, interrupted iteration, FIFO rejection before reading, stale-refresh cleanup, four display widths, keyboard focus, targeted accessibility, and extracted portable-launcher startup and shutdown.

The portable archive was built twice with identical bytes and checked for ZIP integrity. Syntax, documentation links, Git integrity, and release privacy checks were run. GitHub CI repeats Linux/macOS tests, Chromium/WebKit tests, and packaging before publication. Passing regression fixtures do not establish comprehensive browser compatibility, real-world malware detection accuracy, or resistance to every hostile PDF.

See the PDF Inspector guide, validation guide, and changelog for workflow details and limits. The published 2.1.0 tag and archive are unchanged.

Portable artifact

Filename: macos-inspector-2.2.0-macos.zip
SHA-256: 7cb3449bed37ad364eaacfc895e0da41b6cc397c6ea5b555807e63ce1f7c8acf

Verify the ZIP against the matching SHA256SUMS before launching. The package is not Apple-signed or notarized.