v1.0.0 — Initial Release

AI & Cybersecurity Skills v1.0.0

Initial public release of ten evidence-driven security skills for Codex and compatible agents.

AI security

• threat-model-ai-system
• review-prompt-injection
• review-agentic-security
• review-rag-security
• red-team-llm-application
• evaluate-ai-guardrail

Cybersecurity

• review-web-application-security
• review-api-security
• review-cloud-iam
• triage-security-incident

Every skill includes trigger metadata, authorization and safety boundaries, evidence rules, an actionable workflow, a fixed output contract, a quality gate, UI metadata, and an official-source methodology reference.

The repository also includes 40 positive, edge, safety, and non-trigger eval cases; a dependency-free collection validator; validator regression tests; contribution and disclosure policies; and read-only GitHub Actions validation.

Active LLM red-team testing requires explicit skill invocation and explicit target authorization. Framework mappings organize coverage and do not claim certification or compliance.

Release assets

Each skill is available as an independently installable ZIP. Verify downloads with SHA256SUMS. The source archives generated by GitHub contain the complete collection and Apache-2.0 license.