This repository packages the active pieces of an AI-assisted OpenStack review workflow:
- Claude plugin and marketplace metadata
- review agents and the
/teim-reviewskill - review report schema and helper transforms
- Zuul jobs and Ansible roles that run the workflow in CI
The repository started as a generic AI style-guide project. It has since evolved into a review system with a growing internal knowledge base. The runtime pipeline is the primary product, while standards snapshots and derived review knowledge support that pipeline.
The current production path is:
- Zuul checks out this repository as a required project.
ai_review_setupregisters the local marketplace and installs theteim-reviewplugin.ai_code_reviewinvokesteim-review-agentin one Claude session.- The agent orchestrates context extraction, commit summarization, project-guideline extraction, and structured review generation.
- Python helpers turn the JSON report into HTML and Zuul file comments.
Local interactive usage follows the same model through /teim-review.
/plugin marketplace add /path/to/openstack-ai-style-guide
/plugin install teim-review@openstack-ai-style-guide
/teim-reviewLocal output is written to .teim-review/:
zuul-context.mdcommit-summary.mdproject-guidelines.mdreview-report.jsonreview-report.html
The live job entrypoint is teim-code-review, defined in
zuul.d/jobs.yaml and executed via playbooks/teim-code-review/run.yaml.
openstack-ai-style-guide/
├── .claude-plugin/ # Claude plugin and marketplace metadata
├── agents/ # Review orchestration and specialist agents
├── skills/ # Interactive skill entrypoints
├── schemas/ # Structured output contracts
├── tools/ # JSON → HTML and JSON → Zuul helpers
├── roles/ # Ansible roles used by the Zuul workflow
├── playbooks/ # Zuul playbooks
├── zuul.d/ # Jobs, projects, semaphores
├── docs/ # Baseline guides, knowledge overlays, and legacy docs
├── references/ # Canonical external standards snapshots
└── tests/ # Unit and contract tests
The repository now treats review knowledge in three layers:
- Canonical external snapshots:
references/ - Curated baseline and derived knowledge:
docs/ - Runtime behavior and contracts:
agents/,skills/, andschemas/review-report-schema.json
references/ holds canonical snapshots of external coding standards, policy,
and review guidance. docs/quick-rules.md and
docs/comprehensive-guide.md are curated baseline guides distilled from those
references. docs/knowledge/ is the starting point for internal overlays,
examples, and future RAG-oriented review knowledge.
Agents and the schema remain authoritative for review behavior, orchestration, and output contracts.
See docs/review-system-overview.md for the runtime model and docs/archive/README.md for legacy material status.
This restructuring pass preserves the current runtime contracts:
- plugin name:
teim-review - marketplace identity:
openstack-ai-style-guide - skill entrypoint:
/teim-review - orchestrator:
teim-review-agent - output schema:
schemas/review-report-schema.json - local output directory:
.teim-review/ - Zuul jobs:
teim-code-reviewandopenstack-ai-style-guide-lint
Common checks:
python3 -m unittest discover -s tests/unit -p 'test_*.py'
tox -e py3
tox -e linterstox is the main developer path when available. The unit tests also cover
repo contracts such as plugin metadata, skill wiring, and workflow references.
The repo still includes docs/quick-rules.md,
docs/comprehensive-guide.md, a new docs/knowledge/ area, and legacy
checklists/templates material. These serve different purposes:
- baseline review context for the current workflow
- constructed internal knowledge that will grow over time
- legacy support content pending further pruning or relocation
Apache License 2.0. See LICENSE.