Arabic AI Skills v1.2.3

Arabic AI Skills v1.2.3

A final hardening release focused on safer agent behavior, SDK privacy, reproducible packaging, and durable project documentation.

Security and privacy

• Both skills now treat supplied copy, files, and interface artifacts as untrusted data rather than governing instructions.
• The optional Node.js SDK bounds all text inputs before making an API request.
• OpenAI response storage is disabled by default; callers can explicitly opt in with store: true.
• Product context is serialized as data with an explicit instruction boundary.

Reliability

• Distribution ZIP names and contents now derive from the same committed revision.
• Packaging tests verify that both skills are present and large demo media is excluded.
• CI documentation now links to the canonical workflow instead of duplicating action revisions that can become stale.
• Marketing and compatibility language now matches the evidence published in the repository.

Verification

• 10 automated tests pass.
• Node.js 22 and 24 CI matrix.
• TypeScript type checking, Biome linting, schema/plugin/benchmark validation, CodeQL, and Dependabot.
• npm audit: 0 vulnerabilities.

See CHANGELOG.md for details.

Arabic AI Skills v1.2.2

Arabic AI Skills v1.2.2

A hardened public release of two production-focused Arabic agent skills:

• arabic-intelligence — natural Arabic product copy, localization, terminology, and copy audits.
• arabic-uiux-master — Arabic interface writing with RTL, typography, and accessibility guidance.

Highlights

• One-command installation through skills.sh-compatible tooling.
• Codex UI metadata and cross-agent skill packaging.
• 9 automated tests plus Node.js 22/24 CI and CodeQL scanning.
• Dependabot maintenance and pinned GitHub Actions.
• MIT license, security policy, issue templates, and contribution guidance.
• 54-second before/after demo covering GovTech, payments, and permission prompts.

Security note

v1.2.2 removes redundant self-install commands from packaged skill documentation. Gen Agent Trust Hub reports Safe and Socket reports 0 alerts. The remaining Snyk E005 result on skills.sh references the older indexed snapshot and has been requested for re-scan.

See CHANGELOG.md for details.