An automated 8-step content creation pipeline powered by GitHub Copilot custom agents. Provide a technical topic and get a full distribution package: long-form blog, social posts (LinkedIn, X/Twitter, Reddit), and YouTube script — all with consistent visuals and data-driven quality.
- VS Code (or VS Code Insiders) with GitHub Copilot Chat extension
- Python 3.10+ (for visual rendering)
- A GitHub Copilot subscription with custom agents support
python3 -m venv .venv
source .venv/bin/activate
pip install matplotlibEdit content/pipeline-config.md to set:
- Reference URLs — online sources the pipeline will fetch and analyze before writing
- Model preferences — which Copilot model to use per task (informational; you select the actual model in the VS Code picker)
- Output preferences — blog length, target subreddits, YouTube duration
Open the Copilot Chat panel and select your preferred model from the model picker dropdown. Run /configure-model in chat to see all available models with task-specific recommendations.
Full pipeline — type in Copilot Chat:
@content-pipeline Create content about [your topic]
Or use the prompt shortcut:
/new-content-pipeline
The orchestrator will guide you through all 8 steps with quality gates between phases.
Open content/pipeline-config.md and check the Pipeline Status section at the top.
| Status | What It Means | What to Do |
|---|---|---|
not-started |
No active run | Edit config (references, preferences) → run @content-pipeline |
in-progress |
Pipeline is mid-run | Run @content-pipeline — it resumes from the last incomplete step |
completed |
All steps done | Review content → run /archive-content → then start a new run |
blocked |
A step needs attention | Check the step checklist for details, fix the issue, then resume |
- Check status in
content/pipeline-config.md - If previous content exists (
completedorin-progress), run/archive-contentfirst - Edit references, model preferences, and output preferences
- Run
@content-pipeline Create content about [your topic]
- Check the step checklist in
content/pipeline-config.md— checked boxes show what's done - Run
@content-pipeline— the orchestrator reads the checklist and picks up where it left off - No need to re-specify the topic; it's saved in the status section
| Step | Agent | What It Does |
|---|---|---|
| 0 | @content-pipeline |
Fetches and analyzes reference URLs from config |
| 0b | @trend-researcher |
Market intelligence, competitive landscape, data points |
| 1-2 | @content-strategist |
Clarifying questions → strategy doc + outline |
| 3 | @blog-writer |
Long-form blog (~3,000 words) |
| 3b | @visual-renderer |
PNGs, SVGs, and Mermaid diagrams |
| 3c | @quality-reviewer |
Quality audit with fixes |
| 3d | @seo-optimizer |
SEO metadata, keyword optimization, heading structure |
| 4a | @social-strategist |
Cross-platform social distribution strategy |
| 4b | @social-linkedin |
Plain + Unicode formatted LinkedIn posts |
| 5 | @social-twitter |
Tweet thread + standalone summary |
| 6 | @social-reddit |
Markdown Reddit post |
| 7 | @brand-guardian |
Brand consistency audit across all content |
| 7b | @grounded-content-reviewer |
Web-search-grounded fact-checking and gap analysis |
| 8 | @video-scriptwriter |
YouTube script with slide map |
| 9 | @content-repurposer |
Newsletter, slide deck, podcast, infographic |
| 10 | @web-publisher |
Publish blog to GitHub Pages site |
You don't have to run the full pipeline. Use any agent directly:
@blog-writer Write a blog post from content/my-topic-strategy.md
@social-linkedin Adapt content/my-blog.md for LinkedIn
@quality-reviewer Review content/my-blog.md
@visual-renderer Create visuals for content/my-blog.md
@trend-researcher Research market landscape for [topic]
@seo-optimizer Optimize content/my-blog.md for search
@brand-guardian Audit all content for brand consistency
@grounded-content-reviewer Validate content against web sources
@social-strategist Create social distribution plan for content/my-blog.md
@content-repurposer Repurpose content/my-blog.md into newsletter, slides, podcast
Before a content run, add your research sources to content/pipeline-config.md under Reference URLs:
**Industry Reports & Benchmarks:**
- [State of AI 2026](https://example.com/report) — extract market size, adoption rates
**Pricing Pages & Documentation:**
- [OpenAI pricing](https://openai.com/pricing) — per-token costs for GPT-4.1The pipeline fetches these in Step 0 and produces content/reference-brief.md — a synthesized analysis that all downstream agents use for data-backed writing.
| Command | Description |
|---|---|
/new-content-pipeline |
Start a full pipeline run |
/quality-review |
Run quality audit on existing content |
/configure-model |
Discover available Copilot models + recommendations |
/archive-content |
Archive current content and prepare for a new run |
.github/
├── copilot-instructions.md # Workspace-wide rules (tokens, quality, tone)
├── agents/ # 15 specialist agents
├── skills/ # 3 reusable skills
│ ├── visual-rendering/ # PNG/SVG/Mermaid generation
│ ├── unicode-formatting/ # Bold/italic for social posts
│ └── reference-analysis/ # Fetch + synthesize online sources
├── instructions/ # 3 auto-loading instruction files
├── prompts/ # 4 prompt shortcuts
├── ISSUE_TEMPLATE/ # Bug report & feature request templates
├── PULL_REQUEST_TEMPLATE.md # PR checklist
└── CODEOWNERS # Review assignments
content/
├── pipeline-config.md # ← Edit this before each run
├── reference-brief.md # Auto-generated from reference URLs
├── *.md # Blog, social posts, scripts
└── visuals/ # PNGs, SVGs, Mermaid files, renderers
archive/ # Past content runs (max 3 kept)
├── run-YYYYMMDD-HHMMSS/
│ ├── *.md
│ └── visuals/
scripts/
├── archive-content.sh # Archive + rotate content runs
└── Validate-DocsReadme.ps1 # Validate all docs/ subfolders have README.md
agents-and-skills/
└── automation-architecture.md # Detailed architecture documentation
All visuals use a shared token system for consistent branding:
| Token | Color | Usage |
|---|---|---|
| ACCENT | #1f6feb |
Primary blue |
| ACCENT_2 | #0d9488 |
Secondary teal |
| ACCENT_3 | #7c3aed |
Tertiary purple |
| WARN | #dc2626 |
Warning red |
| SUCCESS | #16a34a |
Positive green |
Full token reference: .github/skills/visual-rendering/references/design-tokens.md
Every piece of content must meet these bars:
- Concrete numbers, model names, or benchmarks — no vague claims
- Real pricing data, not placeholders
- At least one case study with before/after metrics
- First-person tone: "sharing my learnings working with customers"
- No corporate/fundraising framing
When you finish a content run and want to start fresh:
Option A — Prompt shortcut:
/archive-content
Option B — Shell script:
./scripts/archive-content.shThis will:
- Show you what will be archived and ask for confirmation
- Copy all content (except
pipeline-config.md) toarchive/run-YYYYMMDD-HHMMSS/ - Clean
content/for a fresh pipeline run - Keep only the last 3 archives — older ones are automatically pruned
- Add a new pipeline step: See the architecture doc
- Change visual style: Edit design tokens
- Adjust quality rules: Edit
.github/instructions/content-quality.instructions.md - Change social formatting: Edit
.github/instructions/social-formatting.instructions.md
Five agents in this pipeline are adapted from the agency-agents project (MIT License) — trend-researcher, brand-guardian, seo-optimizer, social-strategist, and content-repurposer. See individual agent files for source details.
Contributions are welcome! See CONTRIBUTING.md for guidelines, commit conventions, and code standards.
Please read our Code of Conduct before participating.
To report a vulnerability, see SECURITY.md. Do not open public issues for security concerns.
This project is licensed under the MIT License.