The parallel code review process expands from 3 to 5 reviewers to enhance code quality. This introduces `ring:test-reviewer` and `ring:nil-safety-reviewer` to provide deeper analysis of test coverage and null safety patterns, improving overall reliability. The `dev-cycle` workflow is upgraded from a 6-gate to a 10-gate process, incorporating a sophisticated suite of automated testing gates: - Unit Testing (Gate 3) - Fuzz Testing (Gate 4) - Property-based Testing (Gate 5) - Integration Testing (Gate 6) - Chaos Testing (Gate 7) This change establishes a much higher standard for testing and ensures features are more resilient and production-ready. Finally, new agents and commands are added to the PMO and Product plugins to support delivery tracking, design validation, and reporting, further extending the system's capabilities across the development lifecycle. All documentation is updated to reflect these enhancements.
8 KiB
Ring Workflows Reference
This document contains detailed workflow instructions for adding skills, agents, hooks, and other Ring components.
Adding a New Skill
For Core Ring Skills
-
Create directory:
mkdir default/skills/your-skill-name/ -
Write
default/skills/your-skill-name/SKILL.mdwith frontmatter:--- name: your-skill-name description: | Brief description of WHAT the skill does (method/technique). trigger: | - Specific condition that mandates this skill - Another trigger condition skip_when: | - When NOT to use → alternative skill - Another exclusion sequence: after: [prerequisite-skill] # Optional: ordering before: [following-skill] related: similar: [differentiate-from] # Optional: disambiguation complementary: [pairs-well-with] --- -
Test with:
Skill tool: "ring:testing-skills-with-subagents" -
Skill auto-loads next SessionStart via
default/hooks/generate-skills-ref.py
Production Readiness Audit (ring-default)
The production-readiness-audit skill (ring:production-readiness-audit) evaluates codebase production readiness across 27 dimensions in 5 categories. Invocation: use the Skill tool or the /ring:production-readiness-audit command when preparing for production, conducting security/quality reviews, or assessing technical debt. Batch behavior: runs 10 explorer agents per batch and appends results incrementally to a single report file (docs/audits/production-readiness-{date}-{time}.md) to avoid context bloat. Output: 27-dimension scored report (0–270) with severity ratings and standards cross-reference. Implementation details: default/skills/production-readiness-audit/SKILL.md.
For Product/Team-Specific Skills
-
Create plugin directory:
mkdir -p product-xyz/{skills,agents,commands,hooks} -
Add to
.claude-plugin/marketplace.json:{ "name": "ring-product-xyz", "description": "Product XYZ specific skills", "version": "0.1.0", "source": "./product-xyz" } -
Follow same skill structure as default plugin
Modifying Hooks
-
Edit
default/hooks/hooks.jsonfor trigger configuration -
Scripts in
default/hooks/:session-start.sh- Runs on startupclaude-md-bootstrap.sh- CLAUDE.md context
-
Test hook output:
bash default/hooks/session-start.shMust output JSON with
additionalContextfield -
SessionStart hooks run on:
startup|resumeclear|compact
-
Note:
${CLAUDE_PLUGIN_ROOT}resolves to plugin root (default/for core plugin)
Plugin-Specific Using-* Skills
Each plugin auto-loads a using-{plugin} skill via SessionStart hook to introduce available agents and capabilities:
Default Plugin
ring:using-ring→ ORCHESTRATOR principle, mandatory workflow- Always injected, always mandatory
- Located:
default/skills/using-ring/SKILL.md
Ring Dev Team Plugin
ring:using-dev-team→ 10 specialist developer agents- Auto-loads when ring-dev-team plugin is enabled
- Located:
dev-team/skills/using-dev-team/SKILL.md - Agents (invoke as
ring:{agent-name}):- ring:backend-engineer-golang
- ring:backend-engineer-typescript
- ring:devops-engineer
- ring:frontend-bff-engineer-typescript
- ring:frontend-designer
- ring:frontend-engineer
- ring:prompt-quality-reviewer
- ring:qa-analyst
- ring:sre
- ring:ui-engineer
Ring PM Team Plugin
ring:using-pm-team→ Pre-dev workflow skills (8 gates)- Auto-loads when ring-pm-team plugin is enabled
- Located:
pm-team/skills/using-pm-team/SKILL.md - Skills: 8 pre-dev gates for feature planning
Ring TW Team Plugin
using-tw-team→ 3 technical writing agents for documentation- Auto-loads when ring-tw-team plugin is enabled
- Located:
tw-team/skills/using-tw-team/SKILL.md - Agents (invoke as
ring:{agent-name}):- ring:functional-writer (guides)
- ring:api-writer (API reference)
- ring:docs-reviewer (quality review)
- Commands: write-guide, write-api, review-docs
Ring FinOps Team Plugin
using-finops-team→ 3 FinOps agents for Brazilian compliance and cost estimation- Auto-loads when ring-finops-team plugin is enabled
- Located:
finops-team/skills/using-finops-team/SKILL.md - Agents (invoke as
{agent-name}):- finops-analyzer (compliance analysis)
- infrastructure-cost-estimator (cost estimation)
- finops-automation (template generation)
Hook Configuration
- Each plugin has:
{plugin}/hooks/hooks.json+{plugin}/hooks/session-start.sh - SessionStart hook executes, outputs additionalContext with skill reference
- Only plugins in marketplace.json get loaded (conditional)
Creating Review Agents
-
Add to
default/agents/your-reviewer.mdwith output_schema (see AGENT_DESIGN.md) -
Reference in
default/skills/requesting-code-review/SKILL.md:85 -
Dispatch via Task tool:
subagent_type="ring:your-reviewer" -
MUST run in parallel with other reviewers (single message, multiple Tasks)
Pre-Dev Workflow
Simple Features (<2 days): /ring:pre-dev-feature
├── Gate 0: pm-team/skills/pre-dev-research
│ └── Output: docs/pre-dev/feature/research.md (parallel agents)
├── Gate 1: pm-team/skills/pre-dev-prd-creation
│ └── Output: docs/pre-dev/feature/PRD.md
├── Gate 2: pm-team/skills/pre-dev-trd-creation
│ └── Output: docs/pre-dev/feature/TRD.md
└── Gate 3: pm-team/skills/pre-dev-task-breakdown
└── Output: docs/pre-dev/feature/tasks.md
Complex Features (≥2 days): /ring:pre-dev-full
├── Gate 0: Research Phase
│ └── 3 parallel agents: repo-research, best-practices, framework-docs
├── Gates 1-3: Same as simple workflow
├── Gate 4: pm-team/skills/pre-dev-api-design
│ └── Output: docs/pre-dev/feature/API.md
├── Gate 5: pm-team/skills/pre-dev-data-model
│ └── Output: docs/pre-dev/feature/data-model.md
├── Gate 6: pm-team/skills/pre-dev-dependency-map
│ └── Output: docs/pre-dev/feature/dependencies.md
├── Gate 7: pm-team/skills/pre-dev-task-breakdown
│ └── Output: docs/pre-dev/feature/tasks.md
└── Gate 8: pm-team/skills/pre-dev-subtask-creation
└── Output: docs/pre-dev/feature/subtasks.md
Development Cycle (10-gate)
The ring:dev-cycle skill orchestrates task execution through 10 gates: implementation (Gate 0) → devops (Gate 1) → SRE (Gate 2) → unit-testing (Gate 3) → fuzz-testing (Gate 4) → property-testing (Gate 5) → integration-testing (Gate 6) → chaos-testing (Gate 7) → review (Gate 8) → validation (Gate 9). All gates are MANDATORY. Invoke with /ring:dev-cycle [tasks-file] or Skill tool ring:dev-cycle. State is persisted to docs/ring:dev-cycle/current-cycle.json. See dev-team/skills/dev-cycle/SKILL.md for full protocol.
Parallel Code Review
Instead of sequential (100 min)
review1 = Task("ring:code-reviewer") # 20 min
review2 = Task("ring:business-logic-reviewer") # 20 min
review3 = Task("ring:security-reviewer") # 20 min
review4 = Task("ring:test-reviewer") # 20 min
review5 = Task("ring:nil-safety-reviewer") # 20 min
Run parallel (20 min total)
Task.parallel([
("ring:code-reviewer", prompt),
("ring:business-logic-reviewer", prompt),
("ring:security-reviewer", prompt),
("ring:nil-safety-reviewer", prompt),
("ring:test-reviewer", prompt)
]) # Single message, 5 tool calls
Key rule
Always dispatch all 5 reviewers in a single message with multiple Task tool calls.
Related Documents
- CLAUDE.md - Main project instructions (references this document)
- AGENT_DESIGN.md - Agent output schemas
- PROMPT_ENGINEERING.md - Language patterns