Changelog

Plugin 6.1.1

• Add portable OpenAI plugin metadata, guided setup, focused skill interfaces, and an explicit skills-only public package boundary.
• Route from security, failure, and verification state before release intent; distinguish ChatGPT and Codex and fix substring routing collisions.
• Defer detailed lifecycle recipes instead of eagerly loading them.
• Build deterministic, security-gated upload ZIPs with archive regression tests.
• Require one explicit new installer destination and preserve existing installs.
• Document host smoke cases and remaining portal publication facts.

All notable changes to the OmniSkill project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[6.2.0] - 2026-10-29

Added

• Automatic Provenance & Intellectual Property Signing:
  • omni-skill init automatically stamps newly initialized skills with author: Omni Skill, repository: https://github.com/imMamdouhaboammar/omni-skill, metadata, and an ## Attribution & Provenance section.
  • Generates signing headers across operational recipes (usecases/run.md) and specification schemas (references/spec.md).
  • TypeScript compiler (compiler.ts) and Python compilers (compile_spec.py, compile_skill_spec.py) inject provenance signatures and attribution sections into compiled artifacts.
  • Skill and tool scaffolders (init_skill.py, scaffold-repo.ts) stamp generated packages, scripts, and distribution manifests (package.json, marketplace.json, .skills.json, install.sh, bin/cli.js) with generator attribution and repository provenance.
  • Extended frontmatter schemas, parser, and type interfaces (SkillFrontmatter, OmniSkillSpec) to natively support author, repository, and metadata.
  • Added strict attribution invariants to SKILL.md authoring standards across omni-skill, skill-conductor, and skill-authoring.

[6.1.0] - 2026-10-26

Added

• Evidence-Derived Evaluation Harness (omni-skill eval):
  • Implemented real case execution against candidate evidence via -c, --candidate and -r, --responses flags (#1).
  • Added -j, --json flag for machine-readable evaluation outputs and CI pipelines.
  • Fail-closed evaluation: unverified cases without candidate responses return exit code 1 instead of false positive 100% passes.
  • Added response fixture in examples/sql-optimizer/evals/responses.json.
• YAML-Safe Frontmatter Serialization:
  • Replaced raw string interpolation with YAML-safe stringification in TypeScript compiler and Python spec compiler (#9).
  • Handles colons, hashes, single/double quotes, multiline block scalars, and Unicode correctly without malforming YAML frontmatter.
  • Standard-library fallback parsing in portability_validator.py and validate_portability.py supporting quoted strings and block scalars without hard PyYAML dependency.

Fixed

• Accessibility Landmarks & Compliance (WCAG 2.1 AA / Insight Code):
  • Added <main> landmark regions to public/index.html (#88).
  • Added <main> landmark regions and scope="col" attributes to evaluation review templates in skills/omni-skill/assets/eval_review.html and skills/skill-conductor/assets/eval_review.html (#90, #93).
  • Added @media (prefers-reduced-motion: reduce) queries for .skill-card and toggle switches while preserving checked knob positions (#89, #93).
  • Added Playwright browser test coverage in skills/omni-skill/scripts/test_viewer_accessibility.py verifying landmarks, scopes, and reduced-motion rules.

[6.0.1] - 2026-10-25

Fixed

• Hardened packaging security preflight and artifact integrity checks (#2).
• Validated compiler path containment and reference containment against traversal attacks (#7, #8).
• Hardened static validator fail-closed boundary enforcement on unreadable system edge cases (#3).

[6.0.0] - 2026-10-25

Added

• Initial release of OmniSkill v6: Universal Cross-Host Agent Skill Engine & Compiler.
• 15 production sub-skills across engineering discipline and autonomous toolchain suites.
• 7-pass static analysis engine and dynamic agentic router.