Skip to content

joncutrer/aiuc-1-context

Repository files navigation

aiuc-1-context

A repository that systematically fetches, archives, tracks, and analyzes the AIUC-1 standard — and distributes role-based Claude Agent Skills for working with it.

What is AIUC-1?

AIUC-1 (AI Use Case standard 1) is the world's first security, safety, and reliability standard designed specifically for AI agents. Developed with 100+ Fortune 500 CISOs, it covers 51 requirements across six domains:

Domain Description
A. Data & Privacy Input/output data policies, PII protection, IP safeguards
B. Security Adversarial robustness, access controls, endpoint protection
C. Safety Harmful output prevention, pre-deployment testing, monitoring
D. Reliability Hallucination prevention, safe tool calls
E. Accountability Failure plans, vendor due diligence, logging, transparency
F. Society Cyber misuse prevention, catastrophic risk controls

The standard maps to: ISO 42001 · MITRE ATLAS · EU AI Act · NIST AI RMF · OWASP Top Ten · CSA AICM


Claude Agent Skills

This repo ships four installable Claude Agent Skills for AIUC-1 compliance work. Each skill gives Claude a specialized role and embeds the full AIUC-1 spec as a reference document.

Install

# Add this repo's dist/ as a marketplace source
claude plugin marketplace add joncutrer/aiuc-1-context/dist

# Install the skills you need
claude plugin install aiuc-1-assessor@aiuc-1-skills
claude plugin install aiuc-1-auditor@aiuc-1-skills
claude plugin install aiuc-1-implementer@aiuc-1-skills
claude plugin install aiuc-1-advisor@aiuc-1-skills

Skills overview

aiuc-1-assessor — Readiness Assessment

Pre-certification gap analysis. Evaluates an AI system against all applicable AIUC-1 requirements and produces a prioritized remediation plan.

Command Description
/assess Full readiness assessment with gap analysis
/scope Determine which requirements apply to your system
/evidence-checklist Generate a checklist of required evidence artifacts

aiuc-1-auditor — Formal Certification Audit

Conducts formal AIUC-1 audits. Verifies evidence artifacts against controls, checks review frequency compliance, and produces structured findings with a certification recommendation.

Command Description
/audit Formal audit report with certification recommendation
/verify Verify a specific evidence artifact against a control

aiuc-1-implementer — Control Implementation

Engineering-focused guidance for building AIUC-1 controls. Provides concrete technical patterns, policy templates, and evidence artifact generation.

Command Description
/implement Implementation guidance for a specific requirement (by ID)
/draft-policy Draft a policy document that satisfies one or more requirements

aiuc-1-advisor — Strategic Advisory

Strategic compliance consulting. Produces roadmaps, vendor questionnaires, framework crosswalk mappings, and executive briefings.

Command Description
/roadmap Phased compliance roadmap (0–90 days, 90–180 days, ongoing)
/vendor-questionnaire Vendor assessment questionnaire based on AIUC-1 requirements
/crosswalk Map AIUC-1 to ISO 42001, EU AI Act, NIST AI RMF, MITRE ATLAS, etc.
/exec-briefing Executive compliance briefing (leadership, customer, or regulatory)

Repository Structure

aiuc-1-context/
├── skills/                    # Authored skill source files (edit these)
│   ├── aiuc-1-assessor/
│   │   ├── SKILL.md
│   │   └── commands/
│   ├── aiuc-1-auditor/       (same structure)
│   ├── aiuc-1-implementer/   (same structure)
│   └── aiuc-1-advisor/       (same structure)
│
├── dist/                      # Built output — commit and distribute from here
│   ├── .claude-plugin/
│   │   └── marketplace.json
│   └── aiuc-1-{assessor,auditor,implementer,advisor}/
│       ├── SKILL.md
│       ├── commands/
│       └── references/
│           └── aiuc-1-spec.md
│
├── src/                       # Automation scripts
│   ├── fetch_spec.py          # Fetch a quarterly spec version from aiuc-1.com
│   ├── diff_specs.py          # Compare two spec versions
│   ├── fetch_news.py          # Harvest news/research articles
│   ├── build_ai_context.py    # Compile all spec data into AI-optimized context
│   ├── build_skills_dist.py   # Assemble skills bundle into dist/
│   └── run_periodic.py        # Orchestrate all tasks in order
│
└── data/
    ├── spec-versions/         # Raw spec snapshots (one folder per quarter)
    ├── changelog/             # Per-quarter changelogs
    ├── spec-diffs/            # Structured diffs between releases
    ├── ai-context/            # AI-optimized single source of truth
    └── news/                  # Monthly news/research digests

Running the Pipeline

Prerequisites: Python 3.11+ and uv

# Install dependencies
uv sync

# Run everything (fetch → diff → build context → build skills)
uv run src/run_periodic.py

# Or run individual steps
uv run src/fetch_spec.py 2026-Q1      # Fetch a specific spec version
uv run src/build_ai_context.py        # Rebuild the AI context doc
uv run src/build_skills_dist.py       # Rebuild dist/ from skills/
uv run src/fetch_news.py              # Harvest latest news articles
uv run src/diff_specs.py 2025-Q1 2026-Q1  # Diff two versions

Typical cadence:

  • Run run_periodic.py quarterly when a new AIUC-1 release is announced
  • Run fetch_news.py monthly to keep news digests current

Customizing the Skills

  1. Edit the source files in skills/<skill-name>/SKILL.md or skills/<skill-name>/commands/<command>.md
  2. Rebuild: uv run src/build_skills_dist.py
  3. The updated dist/ is ready to install from

Never hand-edit files in dist/ — they are always overwritten by the build script.


How It Works

aiuc-1.com  →  fetch_spec.py  →  data/spec-versions/YYYY-QN/
                                          ↓
                              build_ai_context.py  →  data/ai-context/aiuc-1-context-latest.md
                                          ↓
                              build_skills_dist.py  →  dist/aiuc-1-*/
                                                            ├── SKILL.md  (from skills/)
                                                            ├── commands/ (from skills/)
                                                            └── references/aiuc-1-spec.md  (injected)

The AI context document (aiuc-1-context-latest.md) is LLM-optimized: nav menus, marketing content, and duplicate text are stripped. It contains a requirements table with keywords, controls & evidence, domain descriptions, and framework crosswalk content — all in a single dense document.


License

Apache 2.0

About

Archive, track, and analyze quarterly AIUC-1 releases. Includes role-based Claude Agent Skills for AI compliance assessment, auditing, implementation, and advisory.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages