Skip to content

About

Evidence-first Splunk content inventory, dependency analysis, and cleanup review workflow.

Topics

Resources

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

Splunk Content Hygiene

Content Hygiene

CI

Splunk Content Hygiene is an evidence-first Splunk Enterprise app for inventorying knowledge objects, exploring dependencies, finding ownership and repair concerns, and managing a review workflow before any separate cleanup process begins.

The app keeps customer content read-only:

  • It reads real Splunk data and never falls back to a demonstration dataset.
  • It does not delete, disable, reassign, or rewrite customer content.
  • It stores scan results, review workflow records and retention policies only in app-local KV Store collections.
  • Every finding carries reasons, evidence, and a suggested investigation step.

Version 1.16.0 — Investigation brief

Start each selected-object investigation with four plain-language answers: Why review? What is known? What remains uncertain? What should I check next? Candidates, Review Library and Dependency Explorer share the same cache-only brief, with a suggested evidence-opening action and dedicated CSV/JSON exports. The saved decision stays separate from evidence freshness and app protection. Missing objects and historical snapshots remain unknown, not zero dependencies or proof of removal. Review Library opens changed/unverifiable reviews on their evidence, while explicit tab choices and unsaved drafts survive local navigation.

Explorer’s brief and Changes view follow the inspected object, not an unrelated graph center. The relationship suggestion explicitly centers that identity and opens its unfiltered captured links. No new scan, customer-content write, decision migration, dependency/scoring change or broader permission is required. See the Investigation brief guide.

Release 1.16.0, build 113 passes 1,318 automated checks and pre-cert AppInspect with zero errors/failures. Targeted native Light/Dark workflows cover all three brief views on Enterprise 10.2.5, 10.2.8 and 10.4.4, including existing and historical Library decisions where present. Cross-checks cover 6,892 real cached objects, 14 schemas and 28 JSON / 28 CSV downloads. Original journals, inventory manifests and appearance preferences are preserved. See qualification and limits.

Version 1.15.0 — actionable caller-context diagnostics

Understand why a captured search or dashboard query's initial caller context was retained or withheld. Dependency Explorer and Review Library share a Caller context tab with exact query locations, first-rejection explanations, independent syntax observations, expandable provenance and safe next checks. Filter objects by withheld or unrecorded context, page through bounded details, and export the complete retained facts as JSON or CSV regardless of display filters.

An explicit inventory scan captures the new facts; older snapshots say Not recorded. Views read cached data only. No parser grammar, dependency membership, confidence, scoring, review decisions or removal readiness is broadened. Missing, post-process, unavailable and budget-limited evidence remains explicit uncertainty. See the diagnostic guide.

Release 1.15.0, build 112 passes 1,267 automated checks and pre-cert AppInspect with zero errors/failures. Targeted native Explorer/Library Light/Dark checks cover Enterprise 10.2.5, 10.2.8 and 10.4.4. All 13 schemas validate, and 25 JSON / 25 CSV downloads match cached facts. Existing decisions and native appearance settings are preserved; customer content remains read-only. See qualification and limits.

Version 1.14.0 — context-aware alternative paths

Compare a chosen witness with up to three bounded alternatives, using the existing cached graph. Explorer, In-use Content and Removal Impact keep the original path and check every alternative's caller/app/query context independently. Review Library adds an exact-endpoint Dependency paths investigation without changing decisions.

Inspect readable ordered hops and provenance, focus the selected path in Explorer, then return to the previous map. Dedicated CSV/JSON files preserve every displayed path and its search limits; existing full reports keep their original scope. No new scan, customer-content write or broader source permissions are required. Search is bounded, not exhaustive, and no missing path establishes safe removal. See the investigation guide.

Release 1.14.0, build 111 passes 1,230 automated checks and pre-cert AppInspect with zero errors/failures. All seven pages pass desktop Light/Dark checks on Enterprise 10.2.5, 10.2.8 and 10.4.4: 84 accepted native captures and 32 JSON / 32 CSV download comparisons. Review Library loads and integrity-checks cached edges before investigating. Original decisions, journals, baselines and appearance settings are preserved. See qualification and bounded-search limits.

Selected alternative path in Light mode

Built-in Monitoring Console metadata: one selected dashboard → macro → lookup path, with both exact connectors and all three objects emphasized. This is captured static evidence, not observed execution. View the Dark-mode example.

Version 1.13.0 — navigable dependency maps

Release 1.13.0, build 107 names displayed relationship endpoints and highlights their exact connector and cards on hover or keyboard focus. It adds bounded global/branch expansion, viewport-aware object pages, counted continuation groups and a full-window graph workspace. High-degree expansions no longer require a tall stack: browse/search captured connections and focus both exact endpoints together, then return to the same map. Existing dependency evidence, review decisions and customer content are unchanged. The exact archive passes 1,194 automated checks and pre-cert AppInspect with zero errors/failures. All seven pages pass desktop Light/Dark checks on Enterprise 10.2.5, 10.2.8 and 10.4.4: 120 accepted native captures and 12 JSON / 10 CSV downloads, including full reports while the map shows only a focused pair. See the visualization guide and qualification and narrow-pane limits.

Full-window dependency map in Light mode

Built-in Monitoring Console content, with one connection and both endpoint cards emphasized. View the Dark-mode example.

Version 1.12.0 — context-aware witness paths

Release 1.12.0, build 100 annotates chosen multi-hop investigations using existing cached evidence. Explorer graph details, required-content witnesses, removal impact and CSV/JSON distinguish shared captured caller receipts, app-only evidence, mixed contexts and unverified/bounded checks. Independent queries, panels and branches cannot be combined into a stronger witness.

All potential dependencies remain included. A shared receipt is not observed execution; no shared receipt is not proof a dependency is impossible or an object is unused. Decisions, baselines and collector coverage versions remain unchanged. See the investigation guide. The exact archive passes 1,153 automated checks and pre-cert AppInspect with zero errors/failures. All seven pages pass in Light/Dark on Enterprise 10.2.5, 10.2.8 and 10.4.4, with 72 accepted captures and 28 JSON / 32 CSV comparisons. Existing cached catalogs and decision/policy journals remain unchanged. Positive shared-caller paths are automated-only contracts because the native catalogs do not contain those cases. See qualification and recovery limits.

Version 1.11.0 — caller-supported cross-dimension paths

Release 1.11.0, build 96 connects an earlier field producer under one props dimension to a consumer under another only when one captured search explicitly selects both scopes in the same app context. Common field names, separate searches, dashboard panels and case-only source/sourcetype possibilities are not co-applicability evidence.

Supporting caller receipts retain the exact object, app, query and selector locations. Relationship details provide wrapping, filterable, paginated caller links; the same evidence reaches required-content, impact and CSV/JSON exports. Indexed work and evidence-size ceilings remain visible uncertainty. shared-static-v10 reads v1–v9 baselines without changing decisions or journals. Customer content remains read-only; no searches, regexes, expressions, events or lookup data are executed or sampled to manufacture a link.

The stale extraction-attribution notice is also corrected for newer supported snapshots. See the investigation guide.

The exact archive passes 1,107 automated checks and pre-cert AppInspect with zero errors/failures. All seven pages in Light/Dark, native/cache/schema comparisons, 108 accepted captures and 34 JSON / 32 paired CSV comparisons cover Enterprise 10.2.5, 10.2.8 and 10.4.4. Existing decisions, journals and native preferences are preserved. The available native catalogs contain no qualifying cross-dimension pair: positive semantics and supporting-caller UI remain automated contracts, not manufactured native content. See qualification and recovered VM-control limits.

Native Dark-mode caller-coverage diagnostics

Light-mode diagnostics. These are real captured counts; zero is not proof of absence or complete coverage. Issue #8 remains complete against its finite static-coverage contract; this is a separately bounded accuracy extension, not a claim of runtime closure.

Version 1.10.0 — declared producer-domain compatibility

Release 1.10.0, build 95 completes the bounded shared-static contract for #8. It connects declared field producers across compatible literal, bounded * / ... and explicitly captured default scopes. Relationships explain both scopes and the static compatibility basis. They do not represent observed events or execution. Source/sourcetype domains are case-sensitive; host domains are case-insensitive. Unrelated scope dimensions and unsupported patterns are not guessed.

Indexed candidates, bounded product-state work, phase/namespace/shadow checks and original output provenance keep investigations conservative and scalable. Explorer and relationship details retain native Splunk controls, wrapping and Light/Dark support; compact evidence reaches required-content, impact and exports. shared-static-v9 preserves decisions, journals and v1–v8 baselines unchanged.

See the #8 acceptance ledger for the finite source-family contract and deliberate limits. Completing that contract is not full effective-props, runtime/private/external dependency closure or removal approval. Investigation guide.

The exact archive passes 1,065 automated checks and pre-cert AppInspect with zero errors/failures. Native/cache/schema comparisons, all seven pages in Light/Dark, 106 accepted captures and 34 JSON / 32 paired CSV downloads cover Enterprise 10.2.5, 10.2.8 and 10.4.4. Decisions, policy history and native preferences are preserved. Native producer witnesses are 46 literal-scope links; positive pattern/default semantics and props-specific Library cases remain automated contracts, not manufactured native content. Recovered collector/pagination and emulated readiness checks are disclosed in the qualification notes.

Native Dark-mode producer-domain diagnostics

Light-mode diagnostics. These are captured counts, not demonstration values; zero is not proof of absence.

Version 1.9.0 — bounded props patterns and default fallbacks

Release 1.9.0, build 94 extends dependency investigations to a deliberately small props pattern language: literal ASCII text, * within a path segment and ... across segments, matched against existing positive exact original-event selectors. Explicitly captured shared [default] declarations remain visible as potential fallbacks. No customer regex, search, field expression or lookup data is executed.

Explorer, relationship details and Review Library explain whether the scope is literal, a supported bounded pattern, a default fallback or unsupported. Native stanza observations and conditional same-attribute ordering retain every possible path. Literal priority defaults to 100, pattern priority to 0; a captured default never outranks a specific stanza. Candidate and character-work ceilings leave unassessed paths unknown. No field-name joins across pattern/default scopes are guessed.

shared-static-v8 preserves existing decisions, journals and v1–v7 baselines. Expanded evidence may request explicit re-review, never silently reaffirm a decision. Full effective props, arbitrary PCRE/modifiers, escaped Windows paths, encoded/non-ASCII scopes, private/system/file layers and runtime execution remain outside this slice. Follow-up #8 remains open. See the pattern/default guide and coverage matrix.

Qualified, wrapping props context in native Dark mode

The native capture intentionally retains Unverified instead of inventing an effective winner. Light-mode context.

Build 94 passes 1,024 automated checks and pre-cert AppInspect with zero errors/failures. Native/cache/schema replay, all seven pages in Light/Dark, 103 accepted captures and 29 JSON / 27 paired CSV downloads cover Enterprise 10.2.5, 10.2.8 and 10.4.4. Original decisions, journals and appearance choices are preserved; customer content remains read-only. Native catalogs include 24 supported pattern declarations but no positive pattern-caller or captured-default witness. Those positives and props-specific Library cases have automated contracts, not manufactured native content. Emulated readiness recoveries and the bounded qualification are documented in the compatibility notes.

Version 1.8.0 — conditional cross-stanza props order

Release 1.8.0, build 93 compares the same exact props attribute across captured shared literal stanzas for one supported initial-search conjunction. It distinguishes conditional highest/lower precedence from unverified context, explains source → host → sourcetype ordering and same-category numeric priority/ASCII ties, and retains every possible dependency path, including lower bindings as removal fallbacks.

Explorer, relationship context and Review Library reuse accessible, wrapping, searchable, filtered and paginated details. The compact evidence travels through freshness, impact, required-content reports and CSV/JSON. Checks run only during explicit scans, reuse existing native reads and never execute customer searches or regexes. Existing decisions, history and v1–v6 evidence baselines remain intact under shared-static-v7; a scan never silently reaffirms a decision.

This is not full effective-props evaluation. Separate query branches are never combined; patterns, defaults, private/system/file layers, runtime activation and execution remain unverified. No ordering result establishes safe removal. See the cross-stanza investigation guide and coverage matrix. Follow-up #8 remains open.

Build 93 passes 968 automated checks and pre-cert AppInspect with no errors/failures. Fresh native/cache/schema comparisons, all seven pages in Light/Dark, keyboard drill-through and 29 JSON / 27 paired CSV downloads cover Enterprise 10.2.5, 10.2.8 and 10.4.4. Existing decisions and appearance choices are preserved. Native catalogs contain no cross-stanza host/source collisions; positive category/priority/ASCII ordering and props-specific Library cases have automated contracts, not manufactured native witnesses. See the qualification and limits.

Cross-stanza investigation in native Dark mode

This built-in-metadata capture deliberately shows an unverified comparison: incomplete coverage does not become a highest-binding or safe-removal claim. Light-mode example.

Version 1.7.0 — shared props precedence and inheritance

Release 1.7.0, build 89 cross-checks native rule-level ownership/export scope against the merged value of the same exact props stanza/attribute. Explorer and Review Library distinguish local merged match, inherited merged match, shadowed in this context, and unverified, with reasons, filtering, pagination and a drill-down to a different returned binding. Shadowed paths stay in the graph as possible fallbacks after removal; no execution is inferred.

Checks run only during explicit inventory scans, not page navigation. Values are compared transiently; only compact IDs, status/reason and digests are cached. Read, byte, candidate, fan-out and wall-clock ceilings leave unassessed contexts explicitly unknown. The same evidence travels through impact, required-content, freshness and CSV/JSON. Existing decisions and v1–v5 baselines remain intact under coverage policy shared-static-v6.

This is not full effective-props evaluation: user/private/system inheritance, per-file local/default provenance, cross-stanza priority, runtime activation and execution remain unverified. See the props investigation guide and coverage matrix.

Build 89 passes 907 automated checks and pre-cert AppInspect with no errors/failures. Fresh native catalog/cache/schema comparisons and all seven pages in Light/Dark cover Enterprise 10.2.5, 10.2.8 and 10.4.4. Actual JSON/CSV exports, original decisions, policy history and restored appearance preferences are verified. See the qualification and limits, including automated-only shadowing and props-specific Library positives. Broader dependency follow-up #8 remains open.

Shared props relationship context in native Dark mode

The capture uses built-in native metadata, not demonstration data or execution evidence. Long context details wrap in a persistent, paginated disclosure. Light-mode example.

Version 1.6.0 — workflow and navigation dependencies

Release 1.6.0, build 85 adds shared Workflow Action and Navigation Menu inventory, exact named view/report links and literal internal app/view routes. Static workflow searches distinguish explicit namespaces from inherited runtime contexts; app-scoped bindings are never guessed from the declaration app. Field-token templates and dynamic menus remain unverified. Configured links are not clicks or execution evidence. External/POST links are never followed, and raw definitions are not cached.

Expanded coverage policy shared-static-v5 preserves existing stages, notes, assignees, history and v1–v4 evidence baselines. It can request explicit re-review; a scan does not reaffirm a decision. A new complete scan is required to collect the two catalogs. See the interaction guide and qualification and limits. The coverage matrix separates collection, parsing, resolution, configuration and execution evidence.

Evidence-heavy investigation reads use smaller pages without increasing request timeouts or truncating the verified inventory.

Build 85 passes 862 automated checks, portable packaging and pre-cert AppInspect with no errors/failures. Fresh native catalog/cache/schema checks and targeted light/dark investigations cover Enterprise 10.2.5, 10.2.8 and 10.4.4, including real menu-to-view removal impact, grouped report links and downloaded CSV/JSON. Original decisions, policy history and appearance preferences are preserved. Static workflow-search namespace positives also have synthetic contracts; native examples do not cover every supported syntax. Full effective-props and dynamic runtime closure remain unverified, so #8 remains open.

Declared navigation relationship in native Dark mode

The app's own menu is shown—not a demonstration dataset. Source locations, exact destinations and configuration-versus-execution qualifications are retained. Light-mode example.

Version 1.5.0 — context accuracy and app retention

Release 1.5.0, build 82 adds explicit possible configuration contexts, literal automatic-lookup OUTPUT/OUTPUTNEW/AS mappings, and independent exact-app retention policies for current and future shared content. Individual decisions, notes, assignees, baselines and history are not rewritten. Policy conflicts remain visible; operator-only policy revisions are reversible and exportable.

Build 82 passes 775 automated checks, packaging and pre-cert AppInspect with no errors/failures. Fresh read-only catalog/cache/schema checks, seven-page Light/Dark smoke checks, detailed context/mapping checks and actual report downloads cover Enterprise 10.2.5, 10.2.8 and 10.4.4. Native app-policy saves, role denials, concurrent-edit conflicts, reversible lifecycle and history exports also pass on all three runtimes. Existing individual decisions and user appearance preferences are preserved. See the qualification and limits.

Workflow actions and navigation menus remain outside the published 1.5.0 coverage; version 1.6 adds a bounded subset. Full effective-props evaluation remains unverified and follow-up #8 remains open. See policy behavior.

Exact-app retention scope preview for Content Hygiene's own shared content

The native Light-mode preview distinguishes policy protection from individual decisions. The object count is an example from installed content, not a fixed scope or a demonstration dataset.

Version 1.4.0 — declared extraction outputs

Release 1.4.0, build 79 adds bounded, names-only attribution for inline EXTRACT-* captures and unambiguous search-time transform REGEX, literal FORMAT and DELIMS/FIELDS declarations. Existing Explorer/Library details expose expandable source locations, capture numbers and declared inputs without caching patterns or event values. This is declared provenance, not regex execution or field availability. Unsupported/dynamic cases remain explicit gaps.

New scans use coverage policy shared-static-v4; v1/v2/v3 decisions and captured baselines remain readable and unchanged. A scan never reaffirms a decision. Build 79 passes 680 automated checks and pre-cert AppInspect with zero errors/failures, three reviewed warnings and one manifest skip. Native catalog/ cache and schema comparisons, targeted Light/Dark checks and 22 actual report downloads cover Enterprise 10.2.5, 10.2.8 and 10.4.4. Original review records and captured baselines remain unchanged. See the qualification and limits, including emulated-VM cold-load timeouts and bounded native Library coverage.

Version 1.3.0 — dependency accuracy

Release 1.3.0, build 77 adds shared Event Type and Tag Binding inventory, literal inclusion/exclusion filters, namespace-aware resolution and explicit uncertainty through investigations, required-content reports, review freshness and CSV/JSON. Existing decisions and their captured evidence are preserved; a coverage-policy expansion can require explicit re-review, not automatic reaffirmation.

The bounded field-lineage increment inventories shared Field Extraction, Field Alias, Calculated Field and Field Transform definitions. Ordered REPORT bindings, declared field-input paths and exact event-scope callers are explicitly potential. Existing object details show field names, phases and provenance; reports retain the same qualifications. Customer SPL, expressions and PCRE are never executed to infer lineage, and raw declaration bodies are not cached.

Release build 77 passes 588 automated checks, packaging and pre-cert AppInspect with no errors or failures (three reviewed warnings). Fresh native scans and catalog/cache comparisons cover Enterprise 10.2.5, 10.2.8 and 10.4.4; targeted light/dark checks cover field facts, wrapped graph/context details and the existing review workflow. Real CSV/JSON retains potential-path provenance and partial coverage. Saved decisions and evidence baselines remain unchanged. Earlier candidates 73/74 qualify event/tag coverage only; they do not contain installed field coverage. See the compatibility scope and field-lineage investigation guide.

What it includes

  • Overview reporting for the latest live scan, including searchable, sortable, paginated per-app health composition
  • Cleanup-candidate filtering, evidence inspection, review stages, and CSV/JSON export
  • Bulk review of selected/page/all-filtered candidates with confirmation, existing-decision protection, cancellation and per-object results
  • An in-use/required-content footprint grouped by type, combining current attributable activity and explicit Retain decisions, or following deliberately selected starting objects through their supported outbound dependencies
  • Grouped, expandable directional dependency exploration, multi-step drill-through, candidate-group and review-stage filters, graph-derived removal-impact simulation, ordered remediation guidance, and relationship/impact export
  • Native automatic-lookup rule inventory, configured lookup-definition/file chains, and explicitly potential caller links from exact event selectors
  • Shared event-type/tag inventory and bounded declared extraction, alias, calculated-field and transform lineage, with exact identities and provenance
  • Shared workflow/navigation inventory and bounded declared internal targets, with source locations, exact app resolution and explicit runtime uncertainty
  • A persistent Review Library with triage, investigation, owner-confirmation, eligibility, retain, and blocked stages, with independent review-evidence validity, scoped coverage, change explanations, explicit reaffirmation, conflict-safe saves, recoverable archiving, revision history, and evidence exports
  • Independent exact-app retention policies for current and future shared objects, with explicit scope confirmation, operator-only coordinated revisions, preserved individual decisions, recoverable lifecycle and history exports
  • Full-width selected-object workspaces with investigation actions at the top, focused evidence/decision/history tabs, previous/next review navigation, and unsaved-draft protection
  • Exact dependency targeting with explicit unavailable/filtered-out states and source-scan change notices instead of unrelated fallback objects
  • Ownership reporting and cross-view navigation
  • Bounded and complete live scans with progress, warnings, collector counts, and scan-report export
  • On-demand 30/90/180-day usage-evidence windows for directly attributable saved-search executions and dashboard access, with explicit complete, partial, and unavailable source coverage

The complete scan inventories installed apps, saved searches, dashboards, macros, event types, tag bindings, workflow actions, navigation menus, field extractions, aliases, calculated fields, field transforms, automatic lookup rules, lookup definitions and files, data models, indexes, sourcetypes, and users visible to the initiating Splunk user. Private objects are excluded from the shared cache; access is restricted to explicit hygiene roles.

Install

  1. Download the release archive and SHA256SUMS from the corresponding GitHub release.
  2. Verify the SHA-256 digest.
  3. In Splunk Web, open Apps > Manage Apps > Install app from file and upload the archive.
  4. Open Splunk Content Hygiene > Settings & Scan Status.
  5. Run a bounded scan first, then a complete scan when the permission and connectivity check succeeds.

The support baseline is Splunk Enterprise 10.2.5 or later on a standalone Linux x86-64 search head with KV Store. Release 1.7.0 build 89 has real-data qualification on 10.2.5, 10.2.8 and 10.4.4: fresh bounded/full scans, native catalog/cache and schema comparisons, all seven pages in Light/Dark, shared-props investigations, required-content and removal-impact exports, and the existing Review Library. Saved decisions, policy journals and evidence baselines are preserved. Broader backend permission/concurrency and native policy lifecycle evidence from 1.5.0 remains separately documented. This is bounded static coverage, not execution telemetry or a large-estate benchmark.

Release 1.2.0 (build 72) adds wrapped labels, opaque relationship surfaces, hover/focus previews and full-context dialogs. Targeted real-data UI checks cover Enterprise 10.2.8 and 10.4.4 in both themes. The minimum runtime evidence above is distinct from these final UI checks; see the exact qualification scope in the compatibility policy below.

App 1.1.1 build 67 retains its broader qualification on 10.2.5, 10.2.8 and 10.4.4: all seven pages in light/dark modes, real-data scans, filter recovery and preservation of existing decisions. Targeted native-browser checks include actual 200% zoom, keyboard review/draft flows and operator-assisted VoiceOver announcements. These are bounded checks, not accessibility certification. Classic/modern navigation and native launcher icons retain their earlier runtime qualification. Maintenance patches do not each require separate release qualification without a known regression. See the compatibility policy for the tested matrix and deployment boundaries, and Installation for upgrade and verification details.

Investigation workflow

  1. Start with a complete scan and review any partial-result warnings.
  2. Collect a bounded usage-evidence window and resolve any source-coverage warnings.
  3. Filter Cleanup Candidates by health group, usage evidence, app, type, or review stage.
  4. Inspect inbound and outbound relationships in Dependency Explorer.
  5. Add an object to Review Library, record evidence, assign it, and move it through the appropriate confirmation stage.
  6. Export the filtered evidence for external review or reporting.
  7. Before relying on a saved decision, check Review evidence and Changes since review. Investigate changes, then explicitly reaffirm the evidence; scans never overwrite your decision.

Coverage separates returned configuration, returned ACL, declared outbound references, and potential inbound references. Current means the supported shared static scope matches a recent reviewed baseline—not that no hidden, private, dynamic, or external consumer exists. Removal-impact planning retains its wider, conservative blockers.

After upgrading to 1.5, run a complete scan to collect output mappings, possible configuration contexts, output attribution and the expanded shared event/tag and field catalogs alongside automatic lookup rules. Follow configured definitions and declared field paths in Explorer. May apply, may read and may use links remain explicitly potential through multi-hop reports, impact explanations and exports. New evidence may require re-review; the scan does not change your saved decision or reaffirm its baseline.

Concurrent saves cannot silently replace one another. A conflicting editor keeps its draft and offers comparison with the latest saved revision. Archiving removes only the active Library entry; its decision and revision history remain available.

Screenshots

The 1.1 required-content report keeps explicit selections separate from observed activity and Retain decisions. This capture starts from Content Hygiene's own application metadata: an App is not an app-wide retain rule, so containment does not expand its footprint. The partial-coverage warning remains visible.

Explicit required-content investigation in dark mode

Light-mode report capture starts from the app's own Dependency Explorer dashboard through its Required content report action; neither capture uses demo data.

Bulk review uses a native decision editor, followed by an exact-scope confirmation and per-object results. See the bulk-review workflow for scope, conflict, cancellation, and export behavior.

The 1.0.1 Review Library gives a selected object the full workspace width, places investigation actions beside its title, and separates decisions, evidence, object details, and history into native tabs. These real-data captures show Content Hygiene's own application metadata; customer content, user identity, and lab navigation are excluded. Current compares a captured inventory snapshot with the reviewed baseline, not live content or removal eligibility.

Full-width Review Library workspace in light mode

Full-width Review Library workspace in dark mode

Dependency links retain the exact requested identity. A center-group mismatch offers explicit recovery instead of silently showing another object. This capture uses the app's own protected metadata, outside the broken-candidate group.

Explicit dependency-target recovery in dark mode

The directional map groups the app's real dashboard and lookup consumers by type, with on-demand expansion and a full evidence table alongside the map.

Grouped dependency neighborhood for Content Hygiene itself

“Confirmed eligible” is an investigation conclusion, not an action. Removal or remediation remains outside this app.

Documentation

Build from source

Node.js 22 or newer and Yarn 1.22 are required.

cd app
yarn install --frozen-lockfile
yarn validate

The verified installable archive is written to app/dist/content_hygiene-<VERSION>.tar.gz (currently 1.7.0, build 89). Run yarn release:prepare to perform the complete validation and update SHA256SUMS.

Current limitations

  • The required-content report is a partial shared static footprint, not a complete migration manifest. Automatic lookup callers are possible exact event-scope matches, not observed use. Pattern scopes, other implicit props, uninventoried types, private/dynamic content and ad-hoc consumers remain explicit gaps. See automatic lookup investigations.
  • Declared field-lineage paths do not establish effective values or runtime use. Capture attribution is a bounded names-only subset, not PCRE compilation or matching. Unsupported/dynamic outputs and expressions, cross-scope precedence, alias chaining and EVAL-to-EVAL chaining remain unverified.
  • Bulk review is not atomic or an app-wide retain rule. Existing decisions are skipped by default; completed saves remain if later items fail or are cancelled.
  • Scans run in the initiating browser session and are not scheduled.
  • Each complete-scan collector is capped at 10,000 visible records and reports truncation.
  • Usage evidence is limited to saved-search and dashboard activity directly attributable through _audit and splunk_web_access records visible to the initiating user. Retention, permissions, and unmappable private namespaces are reported rather than treated as inactivity.
  • Dynamic SPL and token references are disclosed but not resolved as broken.
  • Inline dashboard panels are analyzed through their parent dashboard.
  • Review history is application history, not a tamper-proof audit. Authorized direct KV access can alter it. App saves use a server-side coordinator; coordinated writes require a standalone POSIX search head, not SHC or Windows.
  • Journal history grows with accepted edits. Library loading reads compact revision headers and the required decision/evidence documents; history details load in pages. Plan KV capacity and backups for your review volume.

License

Licensed under the MIT License. Copyright © 2026 Leiter Consulting.

About

Evidence-first Splunk content inventory, dependency analysis, and cleanup review workflow.

Topics

Resources

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages