Skip to content

Anuva Python Server Change Workflow

Shared Anuva workflow skills are supplied by installed anuva-development plugin version 0.2.0. Python retains worker, Web-client, Unity-lifecycle, ZeroMQ, upload, monitoring, service-health, and verification rules in AGENTS.md.

flowchart TD
    Request["Repository-local request"] --> Classify["Plugin ownership and risk preflight"]
    Classify --> Routine["Routine"]
    Classify --> Material["Material"]
    Classify --> ProductMain["Product Main required"]
    Routine --> Deliver["Implement, verify, commit, draft PR"]
    Material --> Plan["Preview exact plan"]
    Plan --> Approval["Separate plan approval"]
    Approval --> Deliver
    ProductMain --> Proposal["Exact Product Main intake; no mutation"]
    Deliver --> MergeApproval["Review exact head and request merge approval"]

Repository-local work

Natural or explicit repository-local requests use $anuva-development:anuva-repository-change.

  • Routine: bounded, low-risk, issue-free work. The request authorizes implementation and focused draft-PR delivery, but never merge.
  • Material: repository-owned and issue-free by default. Preview an exact proportionate plan and stop for separate approval.
  • Product Main required: stop before branch or artifact creation for product decisions, shared contracts, multiple implementation repositories, migrations, authentication/authorization, privacy, security, billing, deployment, release coordination, or uncertain ownership.

Run the lightest complete verification required by AGENTS.md. Preserve worker configuration and lifecycle, Web and Unity adapter boundaries, ZeroMQ/job compatibility, upload and monitoring behavior, service health, security redaction, and unrelated user files. Routine work does not require change-folder evidence unless actual complexity or repository instructions require it.

Explicitly tracked work

Use Linear only for a Product Main implementation child or when the user explicitly requests standalone tracking:

  1. Discover Ready work with $anuva-development:linear-list-ready-issues.
  2. Validate an exact issue with $anuva-development:linear-implement-issue.
  3. For a Ready child, approve the exact anuva work start <id> --confirm mutation.
  4. Preview the exact ImplementationPlan and stop for separate approval.
  5. After approval, implement Python-owned scope, maintain required evidence, and create a linked draft PR through anuva pr create.
  6. Review and complete tracked work through $anuva-development:anuva-review-and-complete-pr.

Documentation

Before editing docs, use $anuva-development:anuva-preview-docs to start or health-check the owned preview. After each logical update, run one anuva docs verify --repository current --file <paths...> and then anuva docs links --repository current --file <paths...>. Never accept an external listener or present the machine-specific preview as canonical.

Guardrails

  • Use guarded anuva operations for supported Linear and PR mutations.
  • Do not create GitHub issues, use GitHub Projects, change Linear manually, or merge directly with gh.
  • Workflow selection never approves a material plan or merge.
  • Do not use YAML workflow execution, Codex Goals, Symphony, or autonomous browser testing.