What changed and why
Material changes to the methodology, schemas, content catalog, and public experience are recorded here.
Current versions
- Methodology
task-fit-1.0.2- Assessment input schema
1.0.0- Assessment result schema
1.0.0- Content catalog
content-1.0.2- Output verification methodology
output-verification-1.0.0- Output verification input schema
1.0.0- Output verification result schema
1.0.0
2026-07-29 · Search, trust, and release hardening
- Restored Pagefind search under the deployed Content Security Policy with the narrow WebAssembly permission it requires, clearer failure guidance, and explicit search-asset revalidation.
- Added one-year HSTS, protected public email links from hosting-layer transformation, and excluded the default Pages hostname from indexing.
- Expanded the privacy notice and clarified that citations are document-level unless a stable relevant-passage locator is displayed.
- Corrected two publisher attributions, replaced internal autonomy codes with readable labels, repaired shared guide wording, and made the Reference library and complete mobile navigation easier to discover.
- Made the owner-reviewed public-beta status and outstanding independent review and moderated user-validation work visible beside the tools.
- Updated dependencies, added an advisory gate, and strengthened release verification so an unexpected artifact file fails the manifest check.
2026-07-28 · Content, trust, and safety remediation
- Rewrote the 85-entry knowledge catalog with topic-specific distinctions, worked examples, assumptions, controls, evidence, acceptance thresholds, and reassessment triggers.
- Added direct domain sources, jurisdiction labels, and a visible explanation of what each source supports.
- Named VMCSoft Digital as the accountable owner reviewer and release approver while explicitly recording that independent domain review and moderated target-user validation remain incomplete.
- Advanced Task Fit to
task-fit-1.0.2. Blocked results now return only disposition-specific next-safe-step codes; their JSON and text exports no longer imply that ordinary AI output checks can make the blocked workflow acceptable. - Added content-quality release gates for blocked-workflow language, scenario completeness, duplicate sources, source support, description duplication, and excessive repeated paragraphs.
- Added clearer homepage journeys, contextual tool actions, curated indexes, a staged Verification Builder, compact mobile navigation, and machine-readable owner-review status.
2026-07-28 · Output Verification Builder
- Added a browser-only builder for checks, evidence, approvals, review coverage, reviewer qualifications, and stop conditions.
- Published 30 VMCSoft Digital owner-reviewed fixtures and representative generated, elevated-review, and blocked checklists.
- Added local copy, print, Markdown, and JSON actions without an account, storage, custom telemetry, or actual-output submission.
- Preserved blockers for missing required capabilities, high-impact decision support, and automatic external action.
The owner-reviewed draft identifier is promoted tooutput-verification-1.0.0 for publication without changing deterministic behavior. Input and result schemas are1.0.0.
2026-07-28 · Safety remediation and public launch
- Changed blocked results so they explicitly prohibit piloting the current workflow and require reassessment after redesign.
- Moved preview and human approval to the beginning of external-action controls and added a prominent before-action boundary.
- Clarified that monitoring and rollback do not replace approval before execution.
- Published the owner-reviewed AI Task Fit Check, examples, methodology, and knowledge hub.
This safety patch advanced the methodology fromtask-fit-1.0.0 to task-fit-1.0.1. Input and result schema structures remained 1.0.0.
2026-07-26 · Private-beta candidate
- Added the deterministic five-step AI Task Fit Check with local answer review and calculation.
- Added complete result explanations, controls, rejected alternatives, pilot guidance, evaluation criteria, and escalation guidance.
- Added local progress, explicit fragment sharing, copy, print, Markdown, and JSON exports.
- Added reviewed examples for all seven solution patterns and integrated the approved Use AI Clearly logo, icon, and palette.
- Recorded product-owner implementation review. The moderated-beta and public-launch gates remain open.
Recommendation behavior remains on methodologytask-fit-1.0.0.
2026-07 · Methodology and knowledge foundation
- Approved deterministic disposition semantics, rule effects, risk meaning, autonomy ceilings, and qualified-review behavior.
- Published the initial 85-page source-backed knowledge catalog, editorial policy, source policy, corrections route, and commercial neutrality policy.
- Added the 95-scenario golden corpus and deterministic Task Fit engine.
