By agentrof
A complete, orchestrated software and product development team: analysis, planning, architecture, design system, implementation, review, verification and a containerized environment.
DevOps engineer role. Spawned by software-engineering-team flows to realize the approved architecture as a runnable containerized environment; never auto-triggered.
Code reviewer role. Spawned by software-engineering-team flows after implementation to audit the change and emit a verdict; never auto-triggered.
Adversarial reviewer role for analysis spaces. Spawned fresh-context by the business-analysis entry with one lens or expert profile and scoped inputs; returns structured findings only.
Backend developer role. Spawned by software-engineering-team flows to implement server-side work against approved contracts; never auto-triggered.
Business analyst role. Runs the interactive analysis persona of software-engineering-team flows and grows the topic's analysis space; invoked with explicit inputs, never auto-triggered.
On-demand reorganization of the whole workspace docs vault. Full-width audit, an owner-gated rename batch, deterministic migration, re-rendered generated views and in-session curation, closing on a green re-check. The vault librarian; scoped stewardship at each docs gate stays where it is.
Interactive business analysis. The analyst persona runs a multi-turn discovery conversation and grows one approved analysis space per topic; the space is the precondition every build and design flow stands on.
Knowledge skill for the analysis challenge loop. Loaded by the business-analysis entry and its challenger and expert roles; not user-facing.
Review framework for the software-engineering-team code reviewer. Defines the finding contract, severity taxonomy, verdict rollup, the single evolving finding record, and the three review passes. Loaded by the reviewer agent during review loops, never invoked directly.
The single change gate for the machine-managed project config. Interprets a requested change, validates it against the supported enums, shows the impact, and writes only workspace/config.json on approval.
Modifies files
Hook triggers on file write and edit operations
Uses power tools
Uses Bash, Write, or Edit tools
Based on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A catalog of curated, end-to-end agent teams for agentic coding harnesses (Claude Code native; Cursor and Codex CLI supported). Install a complete team and run at the goal; this is not a parts store.
The first team, software-engineering-team, is an orchestrated software and product
development team: business analysis, planning, architecture, design
system, implementation, review, verification and a containerized
environment, driven through a small set of user-invocable entry skills.
Every team runs on the project-management-office plugin: a
shared operations backbone holding one central database for projects,
epics, stories, machine-generated tasks, work-order state, findings and audit
events, written only through its CLI and recorded deterministically via
hooks. Installing a team installs project-management-office
automatically as a dependency;
/project-management-office:control-tower starts Control Tower and replies with the running URL.
| Plugins | Agents | Entry skills | Knowledge skills |
|---|---|---|---|
| 2 | 12 | 11 | 14 |
Counts above are injected by tools/counts.py; they are never written by
hand anywhere in this repository.
Claude Code (native):
/plugin marketplace add agentrof/agent-marketplace
/plugin install software-engineering-team
The project-management-office backbone installs automatically with the team (a plugin
dependency); no separate step.
Then, inside your project:
/software-engineering-team:setup
Setup bootstraps the project workspace, asks for anything it cannot detect, and points you at the next step.
Cursor (2.5 or later): add this repository as a team marketplace and install both plugins (the backbone is explicit there), then run the setup entry. Codex CLI:
codex plugin marketplace add agentrof/agent-marketplace
then install both plugins, trust the plugin hooks, and follow the numbered sequence in docs/harnesses.md; setup prints the one-time sandbox paste each harness needs and the restart points.
Everything runs through the team's entry skills; agents and knowledge skills stay behind them.
| Entry | What it does |
|---|---|
/software-engineering-team:setup | Bootstraps the workspace and the project config. Idempotent. |
/software-engineering-team:business-analysis | Interactive analysis; produces the approved brief every flow stands on. |
/software-engineering-team:solution-design | Interactive solution architecture: landscape, technology and topology decisions as living documents the team plans and designs against. |
/software-engineering-team:design-system | Creates or updates the design master from picked candidates. |
/software-engineering-team:sketch | Design exploration: divergent single-file previews, no code. |
/software-engineering-team:demo | Pre-sales package: a navigable single-file demo, no code. |
/software-engineering-team:request | Real work. Atomic asks ship as a small PR; everything else runs the backlog path with gates. |
/software-engineering-team:delivery-lanes | The integrator surface for parallel delivery: proposes lanes, opens worktrees the user drives in their own sessions, owns every merge checkpoint. |
/software-engineering-team:configure | The single change gate for the project config. |
/software-engineering-team:build-docs-vault | On-demand reorganization of the whole docs vault: full audit, owner-gated renames, deterministic migration, curated maps and titles. |
/project-management-office:control-tower | Starts Control Tower, the read-only web dashboard over the central database, and replies with the clickable URL. |
A first session usually looks like: setup, then business-analysis
for the first topic, solution-design for the system foundations,
design-system before any screen work, then request.
tools/: validator, counts injector, scaffolder, tests.make check
runs the validator, the counts drift gate and the test suite. CI runs the same command on every push and pull request; a single finding is red.
See CONTRIBUTING.md. The short version: the validator
is the rulebook, make check must be green, and the named anti-patterns
stay out.
MIT
npx claudepluginhub agentrof/agent-marketplace --plugin software-engineering-teamProject Management Office: the shared operations backbone for every team plugin. One central database for projects, epics, stories, tasks with attempt history, dependencies, DoD records, work orders, findings and audit events, written only through its CLI, recorded deterministically via hooks, and viewed through a read-only web dashboard.
Consult multiple AI coding agents (Gemini, OpenAI, Grok, Perplexity, plus codex, antigravity, and grok CLIs when installed) to get diverse perspectives on coding problems
Production-grade engineering skills for AI coding agents — covering the full software development lifecycle from spec to ship.
Evidence-gated AI coding workflow: scan → analyze → plan → TDD → execute → fix → verify → review, powered by Codebase Memory MCP >= 0.9.0 with optional Serena LSP intelligence. Includes blast-radius planning, test/cycle gates, independent review, and Windows Git Bash hook auto-resolution.
Feature development with code-architect/explorer/reviewer agents, CLAUDE.md audit and session learnings, and Agent Skills creation with eval benchmarking from Anthropic.
Lazy senior dev mode. Forces the simplest, shortest solution that actually works: YAGNI, stdlib first, no unrequested abstractions.
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.