openIndu Workflow

Plugin openindu-control-tower from marketplace openindu in openIndu/control-tower. Commands read from the repo's manifest — yesterday.

Skills (23)

• adopt

  Onboard the current openIndu repo to the public workflow plugin and converge duplicate governance assets (principle copies, duplicate governance agents, ghost /launch commands). Use when a sub-repo first introduces the public principles or has a self-hosted principle.md / PRINCIPLE.md copy.

• autopilot

  Run the SDLC pipeline at a chosen autonomy tier — L0 human-gated, L1 feature-branch, L2 verified staging, or L3 verified disposable sandbox. L2/L3 automation requires a documented target and rollback path. Wraps /design with tiered STOP points; it never bypasses human review or action authorization.

• cleanup

  Close out a session — inventory intermediate artifacts, file durable deliverables into the documented docs/scripts trees with references fixed, and delete throwaways only after the user confirms the printed target list.

• codebase-analysis

  Use when analyzing, reverse-engineering, onboarding to, or documenting any large codebase. Triggers on requests like "分析这个项目", "explain the architecture", "trace the call stack", "how does X work", "code review", or when needing to understand an unfamiliar codebase before making changes. Follows a 7

• company-analysis

  Research and analyze a company for a decision-ready profile or recurring daily company-observation column. Use for company research, industry dynamics, competitive position, or daily company analysis. Do not use as a substitute for personalized investment, legal, or tax advice.

• delivery-check

  Generate and check the RULE 11 delivery-pipeline completeness checklist (sub-repo PR → merge → aggregate submodule → image build → infra-deploy PR → kubectl apply). Must run before reporting any openIndu bug fix / feature complete, to prevent "applied but not live".

• design-md

  Adopt a DESIGN.md design-system document for a project — pick from the awesome-design-md collection or generate one for the user's brand. ui-ux-designer then writes design/uiux/DESIGN.md and frontend implements against it, producing visually consistent UI. Use when a feature needs UI work and there'

• design

  Orchestrate the phase-gated 0-to-1 SDLC workflow across relevant role agents, including independent local testing and applicable operations or metrics artifacts. Use for a new capability, not a one-off fix.

• dev

  Implement one accepted openIndu task against its product acceptance criteria on an isolated feature branch, then hand the exact source SHA to /test.

• dev_issue

  Enumerate one openIndu repository's Issues, triage them and run research, product, development and local testing before creating verified green PRs without merging.

• git-commit-push

  Land the current work as a reviewed commit on a feature branch, push, and open a PR under the repository's own rules — attribution, lint gates, English Gitee PRs, Draft under auto-merge, CI green before ready.

• issue-pipeline

  Collect open issues across the authorized GitHub organization scope, classify which require action, challenge recommendations with independent reviewers, and dispatch accepted work through role agents or /autopilot to a green PR. Use for organization issue triage, not a single already-selected issue

• launch

  Bring up the openindu-maintainers team and spawn manager. Use for cross-repo coordination or when manager/arbiter need the full review-dispatch flow.

• memory

  Read/write the team's reusable lessons ledger. Each role's踩坑/技巧/repeated-wins沉淀 to team/memory/<role>/ so agents don't re-derive them. Every agent reads its role memory at startup and appends a lesson after a non-trivial success or failure. Use to seed a role's memory before a task or to record a ha

• pr_merge

  Check and merge eligible PRs in one explicitly selected openIndu repository, with current-head approval, required CI, serial refresh and authoritative read-back.

• principle

  The sole authoritative source of openIndu's 11 AI Agent behavior RULEs. Any openIndu repo must load this before starting any task (RULE 1). Also load when涉及 git push, K8s manifests, Gitee PR, production DB writes, delivery pipeline, agent authorization tiers, or completed-artifact modification.

• prod

  Turn a task's research evidence into a scoped product requirement, prioritized MVP and executable acceptance matrix for development and local testing.

• publish

  Prepare a verified image release as an English infra-deploy Gitee PR, with an independent repository lease, durable publication journal, remote read-back and rollback plan.

• research

  Research competing products, released features, public roadmaps and evidence-backed differentiation, then recommend concrete actions for one openIndu task.

• retro

  Consolidate the mistakes made during the current session into durable lessons — harvest errors and corrections with evidence, dedupe against the lessons store, write entries, and surface governance gaps as proposals only.

• revision-new

  Create a revision record (changelog) in the control-tower repo logging actual changes to principles / route.json / agent definitions / manifest. Mandatory after editing these control-layer assets; without it, changes aren't traceable.

• route

  Query openIndu repo routing — repo URL, type, host (GitHub/Gitee), image, domain, aggregate ownership, K8s namespace. Use to determine "which repo does this change belong to", "what's this service's image/domain", or "GitHub PR or Gitee PR".

• spec-new

  Create a new spec design draft (feature / process improvement / principle evolution) in the control-tower repo, following spec/TEMPLATE.md with a 5-dimension self-check, then submit to arbiter for review. Cross-repo changes, principle edits, and new RULEs must go through this flow first (RULE 5).

Subagents (20)

• arbiter

  Cross-repo arbiter and review officer. Reviews task nature before manager dispatches and decides routing, adjudicates cross-repo principle/contract conflicts, reviews control-tower spec designs (5 dimensions), and grades inspector proposals. Does not modify repo files (except review-decision files i

• architect

  Architect seat for the SDLC pipeline. Owns technology selection, system architecture, technical decisions, ADRs, the technology roadmap, and tech-debt analysis. Consumes the business analysis and product requirements, produces architecture that the build roles implement. Use when a feature needs tec

• backend

  Backend engineer seat. Skills: Spring Boot (Java/Maven/JUnit/JPA) and FastAPI (Python/uv/pytest/SQLAlchemy/Alembic). Consumes architecture + PRD, implements server code, data access, migrations, build, and tests. Use for API, services, ORM, migrations, server-side config.

• bi-analyst

  BI analyst seat. Owns metrics definition, dashboard design, analytics queries, and insight delivery — turning product/business questions into measurable, checkable success signals. Distinct from data (production-SQL safety). Use to define success metrics, build dashboards, or make a requirement meas

• business-analyst

  Business analyst seat for the 0-to-1 SDLC pipeline. Owns market research, competitor analysis, value-hypothesis framing, and business-case definition. Produces the business analysis report that feeds the product manager. Use when a vague idea needs to become a justified, scoped business case before

• codebase-analyst

  Codebase-insight seat. Uses the codebase-analysis skill''s 7-step methodology to reverse-engineer, onboard to, or document any unfamiliar codebase — open-source or internal. Produces a structured architecture understanding (problem → features → components → code → architecture → 4+1 view → markdown

• control-tower

  Maintainer of the control tower itself. Owns the 11 principles, route.json routing, 20 agent definitions, and the public plugin; does the first-pass spec design and submits to arbiter for review. Use for principle/route/agent-prompt changes, new RULE proposals, or drafting a spec.

• data

  Data seat, owning RULE 10 in full. Production-SQL write protection (multi-column WHERE + BEFORE/AFTER + rowcount-gated conditional commit), migration scripts, dirty-data repair, data quality. Produces plans; a human executes. Use for any production DB UPDATE/DELETE or data repair.

• edge

  Edge engineer seat. Skills: Rust / Cargo / field-protocol drivers (Modbus etc.). Implements edge gateways, protocol drivers, edge alarms, data transforms. Protocol errors fail silently as wrong data — highest risk density. Use for industrial protocols, register maps, edge process stability.

• frontend

  Frontend engineer seat. Skills: React 19, Vue 3 / UniApp, Vite, pnpm, Tailwind. Consumes ui-ux specs + architecture, implements frontend code, components, state, build config, and tests. Use for UI, pages, components, routing, styling, frontend build, or mini-program work.

• inspector

  Proactive read-only inspector. Scans all openIndu repos for principle consistency, CI completeness, push protection, env-var alignment, image-tag drift, K8s-manifest overreach, credential leaks, and design-doc currency. Produces prioritized improvement proposals for manager. Does not modify any file

• manager

  Team manager seat. Classifies user tasks, routes them through arbiter, dispatches to the right SDLC role agents, tracks progress, resolves ordinary workflow exceptions within policy, and checks RULE 11 delivery completeness before handoff. Use when a task spans multiple repos/roles, needs coordinati

• ops

  Operations engineer seat. Owns day-2 operations: server/database health, logs, config validation, incident triage, runbooks, and capacity. Distinct from release (deploy pipeline) — ops keeps running systems healthy. Use for server checks, log analysis, config validation, incident diagnosis, or runbo

• product-manager

  Product manager seat for the SDLC pipeline. Owns requirements clarification, user stories, acceptance criteria, PRDs, cross-repo API contracts, and design reviews. Consumes the business analysis, produces PRDs that the architect and build roles implement. Use when a vague request needs to become an

• release

  Release-pipeline seat. Owns RULE 11's six stages — sub-repo PR → merge → aggregate submodule → image build → gitops manifest → cluster apply — plus the aggregate repo's docker-compose local stack and all K8s manifests in the gitops repo. Use for K8s YAML changes, image-tag switches, Gitee CLI PR cre

• reviewer

  Pre-review seat before human review. Reviews PRs/changes across RULE 5.2's four lenses (correctness, security, maintainability, architectural consistency), producing structured review material for the human reviewer. Has NO approval authority — L2+ changes still need a human. Use for PR pre-review a

• security

  Security seat, covering RULE 5.4 in full. Credential governance, least-privilege review, untrusted-input handling (prompt injection), dependency vulns, external exposure surface (ports/auth/IP allowlist). Engage immediately on credential leak or privilege escalation.

• station-control

  Station & control engineer seat. Skills: C# / .NET / MSBuild — station apps (motion/vision/scan/dispense/laser), edge controllers, upstream .NET protocol-library contributions. The ONLY seat driving physical devices — default L3. Use for C# code, csproj/sln, MSBuild, .NET test projects.

• test

  Test engineer seat. Owns the regression-judgment capability for the org — the RULE 2 admission gate. Does NOT chase test counts; judges whether tests can support regression decisions (repeatability, feedback latency, credibility, coverage truth). Use to add tests, assess agent admission, or fix fail

• ui-ux-designer

  UI/UX designer seat for the SDLC pipeline. Owns interface design, interaction prototypes, the design system, accessibility, and motion. Produces UI specs that frontend implements. Use when a feature needs interface design, interaction flows, a design-system component, or an accessibility review.

Hooks (1)

• PreToolUse

  node "${CLAUDE_PLUGIN_ROOT}/hooks/block-push-main.mjs"

Similar plugins