artifact-orchestration
Orchestrate multi-agent artifact generation using the Primary Author, Parallel Reviewers, Synthesizer, Archive pattern
- License 许可
- MIT
- 所属仓库
- jmagly/aiwg
- 最近更新
- 2026-11-12
- Stars
- ★ 229
- Forks
- 36
01 前言 Front Matter
- name
- artifact-orchestration
- namespace
- aiwg
- platforms
- [ "all" ]
- description
- Orchestrate multi-agent artifact generation using the Primary Author, Parallel Reviewers, Synthesizer, Archive pattern
03 SKILL.md 正文
artifact-orchestration
Orchestrate multi-agent artifact generation with the Primary Author → Parallel Reviewers → Synthesizer → Archive pattern.
Triggers
Alternate expressions and non-obvious activations (primary phrases are matched automatically from the skill description):
- "SAD" → Software Architecture Document
- "software architecture document"
- "architecture document"
- "ADR" → Architecture Decision Record
- "generate [artifact]" for named SDLC artifacts (test plan, deployment plan, requirements)
Purpose
This skill implements the core AIWG multi-agent documentation pattern that supports artifact quality through:
- Single primary author for consistency
- Parallel expert reviews for coverage
- Synthesis of feedback for conflict resolution
- Formal archival with metadata tracking
Behavior
When triggered, this skill orchestrates:
-
Artifact identification:
- Parse requested artifact type
- Map to template and primary author
- Identify required reviewers
-
Workspace setup:
- Create
.aiwg/working/{type}/{name}/ - Initialize metadata via
artifact-metadataskill - Load relevant template via
template-engineskill
- Create
-
Primary author dispatch:
- Launch primary author agent with template
- Provide context (requirements, existing artifacts)
- Receive draft v0.1
-
Parallel review dispatch:
- Launch 3-5 reviewers via
parallel-dispatchskill - Commit each reviewer's initial answer before exposing others' judgments; retain dissent and criteria/check references
- Agreement is not proof, including same-vendor and cross-vendor agreement. Consequential claims require an executable check or accountable human acceptance. Prefer independent voting, or structured review with diverse initial answers and calibrated confidence, over plain debate. Record observed evaluator model/vendor/family, acceptance criteria reference, generator/evaluator disagreements and a re-review link after material fixes. Do not claim reduced human-review time. See
$AIWG_ROOT/docs/decision/evaluator-calibration.mdfor calibration and REF-2922, REF-2923, REF-2919/2921 public sources. - Collect structured feedback
- Launch 3-5 reviewers via
-
Synthesis:
- Invoke Documentation Synthesizer agent
- Merge feedback and resolve conflicts
- Generate final artifact
-
Archive:
- Move to
.aiwg/{category}/ - Update metadata to baselined
- Update artifact index
- Move to
Artifact Configuration
Software Architecture Document (SAD)
artifact: software-architecture-document
template: analysis-design/software-architecture-doc-template
primary_author: architecture-designer
reviewers:
- security-architect
- test-architect
- requirements-analyst
- technical-writer
output: .aiwg/architecture/sad.md
Test Plan
artifact: test-plan
template: test/test-plan-template
primary_author: test-architect
reviewers:
- security-auditor
- requirements-analyst
- devops-engineer
output: .aiwg/testing/test-plan.md
Deployment Plan
artifact: deployment-plan
template: deployment/deployment-plan-template
primary_author: deployment-manager
reviewers:
- devops-engineer
- security-architect
- reliability-engineer
- support-lead
output: .aiwg/deployment/deployment-plan.md
Requirements Specification
artifact: software-requirements-spec
template: requirements/srs-template
primary_author: requirements-analyst
reviewers:
- domain-expert
- system-analyst
- test-architect
- architecture-designer
output: .aiwg/requirements/srs.md
Threat Model
artifact: threat-model
template: security/threat-model-template
primary_author: security-architect
reviewers:
- privacy-officer
- architecture-designer
- devops-engineer
output: .aiwg/security/threat-model.md
Workflow Sequence
┌─────────────────────────────────────────────────────────┐
│ 1. SETUP │
│ • Create workspace: .aiwg/working/{type}/{name}/ │
│ • Initialize metadata.json │
│ • Load template │
└─────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────┐
│ 2. PRIMARY AUTHOR │
│ • Dispatch: architecture-designer (for SAD) │
│ • Input: template, requirements, context │
│ • Output: drafts/v0.1-primary.md │
└─────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────┐
│ 3. PARALLEL REVIEW (single message, multiple agents) │
│ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │
│ │ security- │ │ test- │ │ requirements-│ │
│ │ architect │ │ architect │ │ analyst │ │
│ └──────┬───────┘ └──────┬───────┘ └──────┬───────┘ │
│ │ │ │ │
│ ▼ ▼ ▼ │
│ reviews/security.md reviews/test.md reviews/req.md│
└─────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────┐
│ 4. SYNTHESIS │
│ • Dispatch: documentation-synthesizer │
│ • Input: draft + all reviews │
│ • Output: synthesis/final.md │
│ • Conflicts: document with rationale │
└─────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────┐
│ 5. ARCHIVE │
│ • Move to: .aiwg/{category}/{artifact}.md │
│ • Update metadata: status=baselined, version=1.0.0 │
│ • Archive workspace to: .aiwg/archive/{date}/ │
└─────────────────────────────────────────────────────────┘
Usage Examples
Generate SAD
User: "Create the Software Architecture Document"
Skill orchestrates:
1. Setup workspace for SAD
2. Dispatch architecture-designer with SAD template
3. Launch parallel reviews (security, test, requirements, writer)
4. Synthesize feedback
5. Archive to .aiwg/architecture/sad.md
Output:
"SAD generation complete.
Version: 1.0.0
Status: Baselined
Location: .aiwg/architecture/sad.md
Reviews:
✓ Security Architect: Approved with suggestions
✓ Test Architect: Approved
✓ Requirements Analyst: Approved
✓ Technical Writer: Approved (minor edits)
Conflicts resolved: 2
- Auth approach: Chose JWT per security recommendation
- Caching strategy: Balanced performance vs complexity"
Generate with Context
User: "Generate deployment plan for production release"
Skill uses project-awareness context:
- Phase: Construction
- Environment: AWS
- Compliance: SOC2
Tailors template and reviewer focus accordingly.
Configuration
Custom Artifact Types
Add custom artifact configurations in .aiwg/config/artifacts.yaml:
artifacts:
custom-api-doc:
template: custom/api-documentation
primary_author: api-documenter
reviewers:
- security-auditor
- api-designer
- technical-writer
output: .aiwg/architecture/api-docs/
Review Depth
review_config:
depth: standard # minimal, standard, comprehensive
timeout: 300 # seconds per reviewer
require_unanimous: false # or true for critical artifacts
Integration
This skill uses:
parallel-dispatch: For launching reviewer agentsartifact-metadata: For tracking artifact lifecycletemplate-engine: For loading and instantiating templatesproject-awareness: For context about current phase and state
Error Handling
Reviewer Timeout
If a reviewer times out:
- Mark review as "incomplete"
- Continue with other reviews
- Flag in final synthesis
- User can re-run partial review
Conflicting Reviews
When reviewers disagree:
- Document both positions
- Apply priority rules:
- Security concerns take precedence
- Compliance concerns take precedence
- Domain expert breaks ties
- Record both positions and the criteria/check or human acceptance reference in synthesis; precedence and consensus do not prove correctness
- Re-review material fixes with fresh source bindings and link the earlier review
Missing Template
If template not found:
- List similar templates
- Offer to proceed with generic structure
- Or abort with guidance
Output Locations
| Artifact | Output Path |
|---|---|
| SAD | .aiwg/architecture/sad.md |
| ADR | .aiwg/architecture/adr-{number}.md |
| Test Plan | .aiwg/testing/test-plan.md |
| Deployment Plan | .aiwg/deployment/deployment-plan.md |
| SRS | .aiwg/requirements/srs.md |
| Threat Model | .aiwg/security/threat-model.md |
References
- Multi-agent pattern: docs/multi-agent-documentation-pattern.md
- Templates: templates/
- Agent definitions: agents/
04 同仓库更多 Skills
host-hardening
Read-only system hardening baseline and plan-first host hardening recommendations for SSH, Docker bind addresses, firewall persistence, datastores, exporters and audit retention.
exposure-analysis
Read-only firewall analysis and effective host exposure reconstruction for Docker published ports, UFW bypass, stale rules, CDN scope, IPv6 asymmetry and external probe allowlists.
host-hardening
Read-only system hardening baseline and plan-first host hardening recommendations for SSH, Docker bind addresses, firewall persistence, datastores, exporters and audit retention.
exposure-analysis
Read-only firewall analysis and effective host exposure reconstruction for Docker published ports, UFW bypass, stale rules, CDN scope, IPv6 asymmetry and external probe allowlists.
fixture-skill
Exercise rule delivery.
pi-local
Pi-local duplicate-precedence fixture.