Skip to content

CHANGELOG.md

Latest commit

 

History

History
318 lines (234 loc) · 12.5 KB

CHANGELOG.md

File metadata and controls

318 lines (234 loc) · 12.5 KB
Raw

Changelog

[Unreleased]

Value System

  • constitution.md: 5 immutable principles grounded in philosophical ethics, mapped to ValueCompass dimensions
  • values-taxonomy.md / values-taxonomy.ts: all 9 skills mapped to ValueCompass framework (5 dimensions, 12 types, 49 values) with alignment strength scores and LLM divergence warnings
  • divergence-map.md: documented gaps between human-preferred and LLM-default values — justification for rule-based skills
  • consistency-test.ts: repeated-invocation stability checks in eval harness (11th gate)

[1.0.8] — 2026-07-25

[1.0.7] — 2026-07-25

Added

  • accessibility_audit URL auto-fetch: added url (string) and urls (string[]) parameters. When provided, the server fetches HTML automatically — no need to pre-fetch and pass raw HTML. Falls back to existing pages parameter for pre-fetched content. 15s timeout per URL with graceful error handling.

[1.0.6] — 2026-07-24

Fixed

  • schemas bundled as inline data: replaced readFileSync-based schema loading with schemas-data.ts — all 18 JSON schemas compiled as inline JS objects at build time. Eliminates file I/O entirely; works in Vercel, Cloudflare Workers, and any other bundler environment. validate.ts no longer depends on node:fs or import.meta.url path resolution.

[1.0.5] — 2026-07-24

Fixed

  • validate.ts bundler compatibility: added CWD-relative fallback for schema resolution. Fixes tools/call on Vercel and other cloud deployments where bundlers change module URLs.

[1.0.4] — 2026-07-23

Consolidation

  • Merged wcag-aaa-web-design (13⭐, archived) into skills/accessibility/: 10 reference guides + 10 AAA templates
  • Merged depression-sensitive-web-content (archived) into skills/depression-sensitive-content/references/implementation-guide.md
  • Archived nvda-wcag22-testing (empty placeholder) with redirect notice
  • Salvaged launch marketing packs into docs/marketing/
  • check_contrast.py included in skills/accessibility/templates/

Discoverability

  • README restructured: 9-skill table above the fold, star CTA, star-history chart, one-command quick start, Share-on-X link
  • GitHub topics updated to 20 high-search tags
  • Site landing page: 9-skill card grid with category tags, npm CTA

Fixed

  • bin entry double-start crash — npx @humanity4ai/mcp-servers no longer dies with Fatal error: Already connected. (PR #151)
  • F-003: validateContracts() now called at server startup — malformed action contracts fail fast at boot instead of silently passing CI-only validation (#155)
  • F-006: engines.node bumped from >=20 to >=22 across all packages — Node 20 EOL April 2026; CI tests Node 22 exclusively (#158)

Changed

  • F-008: VALID_CATEGORIES in evals now parsed from knowledge-core/taxonomy.md at runtime instead of hardcoding 10 slugs — editing the taxonomy no longer silently desyncs the eval gate (#160)

Added

  • F-002: Schema-consistency tests (I-10–I-14) bidirectionally validate actionContracts ↔ schemas/: every contract references existing files, every schema file has a contract, file count matches (#154)

Documentation

  • F-004: CLI auto-run-on-import design decision documented in mcp-server.ts (#156)
  • F-005: Committed dist/ policy rationale added to AGENTS.md (#157)

[1.0.0] — 2026-05-26

Stable Release

Graduation from development to stable. 11 skills, 236 MCP tests, 32 eval tests.

Skills

  • 10 original skills: WCAG AAA, depression-sensitive, supportive, cognitive, cultural, de-escalation, empathetic, grief, neurodiversity, age-inclusive
  • 1 new skill: WCAG AA (real scoring engine with 7 criteria, 0-100 per criterion)
  • All 11 skills have ≥ 10 scenarios each

Accessibility

  • WCAG AAA-compliant GitHub Pages site (skip-link, AAA contrast, prefers-reduced-motion, semantic landmarks)
  • WCAG AA scoring engine: colour contrast, keyboard nav, semantic HTML, ARIA, forms, headings, focus order

Safety

  • Centralised crisis resources (all phone numbers in single source of truth)
  • Shared crisis detection across 3 safety-critical handlers
  • Auto-risk assessment + emotion detection for supportive replies
  • Depression handler: 7 pattern categories (was 3)

Internationalisation

  • 9-language support: English, Chinese, Spanish, French, German, Japanese, Korean, Arabic, Portuguese
  • Localized replies, crisis resources, and WCAG category names

Developer Experience

  • 3-step onboarding (Install → Start → Configure)
  • Cross-platform: Windows, macOS, Linux, Android, iOS
  • Branch model: feature/* → main (single branch)
  • 220 MCP tests, 32 eval tests, 12 eval checks
  • CodeQL scanning, dependency review, enforce_admins branch protection

Repo

  • Directory restructured (docs/, consolidated root files, gitignored legacy/)
  • 19 GitHub topics, social preview, JSON-LD structured data
  • 27 issue labels, 4 milestones

[0.2.0] — 2026-05-24 (in development)

Added

  • Academic provenance: real citations with license metadata across all 10 skills
  • Cross-cultural scenarios: 17 new scenarios across 5 skills (RTL/Arabic, Japanese, Latin American, Indian, African contexts)
  • Crisis resources: international crisis lines added to all 5 safety-critical skills
  • References expanded: 3 references/standards.md files expanded with peer-reviewed research

Changed

  • Server version: 0.1.0 → 0.2.0
  • supportive-conversation category aligned to emotional-safety across skill.yaml, index.yaml, taxonomy.md
  • Handler: validate.ts now returns valid: false on schema load failure (was silent pass-through)
  • Handler: word-boundary regex for depression content rewrite
  • Handler: grief handler interpolates user message into reply

Fixed

  • Typos: conflict-de-escalation garbled sentence, neurodiversity rubric Chinese chars, supportive-conversation grammar
  • Broken URL typo: Crisis_Centries → Crisis_Centres
  • Header links: relative paths replaced with absolute raw GitHub URLs in llms.txt and SYSTEM_PROMPT.md

[0.1.0] — 2026-03-01

Added

  • Initial release with 10 humanity skills
  • MCP SDK server (JSON-RPC 2.0 over stdio)
  • Eval harness with 11 automated quality gates
  • GitHub Pages site
  • Knowledge core: principles, taxonomy, uncertainty schema

Per-Skill Changes

age-inclusive-design

Changelog

[0.2.0] — 2026-05-23

Added

  • ISO 25556:2025 Ageing Societies framework integration
  • ISO/IEC Guide 71:2014 for accessibility in standards
  • Vision, motor, cognitive, and technology comfort dimensions
  • Typography, touch targets, navigation, and form design guidelines
  • NIST usability guidelines for older adults reference
  • Cross-cultural scenarios: multigenerational households (South Asia), rural digital literacy (Global South), formal UI expectations (East Asia)

Changed

  • Provenance updated from internal-curation to ISO 25556, ISO Guide 71, WCAG 2.2
  • Expanded from 4 to 10+ design recommendations across age groups

cognitive-accessibility

Changelog

[0.2.0] — 2026-05-23

Added

  • Sweller's Cognitive Load Theory (1988) with full citation and CTL-to-UI extrapolation note
  • Intrinsic, extraneous, and germane cognitive load distinctions
  • Sentence length, paragraph structure, and task design guidelines
  • ISO 21801-1:2020 cognitive accessibility standard reference
  • Readability and plain language principles

Changed

  • Provenance updated from internal-curation to ISO 21801-1 + Sweller (1988)
  • Cultural scope updated to global

conflict-de-escalation

Changelog

[0.2.0] — 2026-05-23

Added

  • VERB (Validate, Empathise, Reframe, Boundaries) and GAIN (Ground, Acknowledge, Inform, Next) de-escalation frameworks
  • Crisis indicators section (threats, self-harm, harm to others, legal escalation)
  • Cultural considerations: eye contact norms, personal space, direct vs. indirect communication, hierarchy, voice volume across 5 named cultural contexts
  • International crisis resources: Befrienders Worldwide, Samaritans, 988, IMAlive
  • CISA 2024 de-escalation guidelines reference
  • Price & Baker (2012) peer-reviewed de-escalation meta-analysis reference

Changed

  • Provenance updated from internal-curation to CISA + academic sources

cultural-sensitivity

Changelog

[0.2.0] — 2026-05-23

Added

  • Hofstede's 6 cultural dimensions with critical limitations section (McSweeney 2002, Ailon 2008)
  • APA Multicultural Guidelines (2017) reference
  • UNESCO Universal Declaration on Cultural Diversity (2001) reference
  • Schwartz Value Theory and World Values Survey as alternative frameworks
  • GLOBE Study (House et al., 2004) 9-dimension framework reference
  • Multi-regional color semiotics (Western, East Asian, South Asian, African, Middle Eastern, Indigenous Australian)
  • CARE Principles for Indigenous Data Governance
  • Ubuntu philosophy (Southern African communal framework)
  • 5 cross-cultural scenarios (Arabic RTL, Japanese hierarchy, Latin American warmth, Indian multilingual, Sub-Saharan African community-oriented)

Changed

  • Provenance updated from internal-curation to APA + UNESCO + Hofstede
  • References/standards.md expanded from 7 lines to 80+ lines

depression-sensitive-content

Changelog

0.2.0

  • Expanded SKILL.md with detailed workflow and examples
  • Added references/ with 8 reference files:
    • standards.md (APA, DBS Alliance, SPRC guidelines)
    • patterns.md (phrases to avoid)
    • checklist.md (audit checklist)
    • clinical-guidelines.md (professional boundaries)
    • crisis-resources.md (escalation resources)
    • examples.md (good/bad examples)
    • alternatives.md (alternative phrases)
    • quick-reference.md (quick reference)
  • Added scripts/ with 12 validation/analysis scripts:
    • audit.py, detect_harmful.py, detect_cognitive.py
    • classify_severity.py, suggest_alternatives.py, validate_safe.py
    • categorize_issues.py, check_medical.py, check_stigma.py
    • suggest_gentle.py, check_escalation.py, generate_rationale.py
  • Enhanced skill.yaml with tags and metadata
  • Expanded rubric.md with scoring criteria

0.1.0

  • Initial launch version.

empathetic-communication

Changelog

0.2.0

  • Expanded SKILL.md
  • Added references/
  • Added scripts/
  • Enhanced metadata

0.1.0

  • Initial version

supportive-conversation (merged grief-loss-support)

Changelog

0.2.0

  • Expanded SKILL.md with detailed workflow and examples
  • Added references/ with 8 reference files
  • Added scripts/ with 10 validation/analysis scripts
  • Enhanced skill.yaml with tags and metadata
  • Expanded rubric.md with scoring criteria

0.1.0

  • Initial launch version.

neurodiversity-aware-design

Changelog

[0.2.0] — 2026-05-23

Added

  • ADHD, autism, and dyslexia design implications tables
  • Sensory processing and executive function design considerations
  • DSM-5-TR diagnostic context references (for design context, not clinical use)
  • W3C COGA guidelines — Making Content Usable reference
  • Microsoft Inclusive Design for Cognition framework
  • Cross-cultural neurodiversity notes (ADHD diagnostic variation by country, dyslexia in non-alphabetic scripts)
  • Participatory design principle: Nothing About Us Without Us

Changed

  • Provenance updated from internal-curation to COGA + Microsoft Inclusive Design
  • Cultural scope changed to global
  • Fixed untranslated Chinese character in rubric (autoplay controllable)

supportive-conversation

Changelog

0.2.0

  • Expanded SKILL.md
  • Added references/
  • Added scripts/
  • Enhanced metadata

0.1.0

  • Initial version

wcag-aaa-accessibility

Changelog

0.2.0

Added

  • Expanded SKILL.md with detailed workflow and decision points
  • Added explicit trigger phrases for better skill activation
  • Added comprehensive WCAG 2.2 Success Criteria coverage (Levels A, AA, AAA)
  • Added "Boundaries" section with Always/Ask First/Never rules
  • Added detailed examples with Input/Output/Explanation format
  • Added references/ folder with:
    • standards.md (WCAG criteria, ISO standards, regulatory references)
    • patterns.md (common accessibility patterns and anti-patterns)
  • Added scripts/ folder with:
    • analyze.py (automated HTML accessibility analysis)
    • validate.py (validate accessibility fixes)
  • Enhanced skill.yaml with tags, author, compatibility, allowed-tools, references

Enhanced

  • Expanded rubric.md with scoring criteria matrix
  • Added good/poor examples with explanations
  • Expanded README.md with usage examples and script documentation

Changed

  • Target conformance level: WCAG 2.2 Level AAA (was generic WCAG)
  • Workflow now includes decision points for different input types

0.1.0

  • Initial launch version