get-content-done

<div align="center">

# Get Content Done

**Sprint planning, two-gate quality, and performance feedback loops for content — powered by Claude Code skills.**

[![Claude Code](https://img.shields.io/badge/Claude_Code-Skills-555?style=for-the-badge&labelColor=D97757&logo=claude&logoColor=white)](https://docs.anthropic.com/en/docs/claude-code)
[![LinkedIn](https://img.shields.io/badge/LinkedIn-Rebecca%20Rae%20Barton-555?style=for-the-badge&labelColor=0077B5&logo=linkedin&logoColor=white)](https://linkedin.com/in/rebeccaraebarton)
[![X](https://img.shields.io/badge/X-@rebeccarae-555?style=for-the-badge&labelColor=000000&logo=x&logoColor=white)](https://x.com/rebeccarae)
[![GitHub stars](https://img.shields.io/github/stars/thatrebeccarae/get-content-done?style=for-the-badge&logo=github&color=121011)](https://github.com/thatrebeccarae/get-content-done/stargazers)
[![License](https://img.shields.io/badge/License-MIT-555?style=for-the-badge&labelColor=353535)](LICENSE)
[![Clone](https://img.shields.io/badge/Clone-git%20clone-555?style=for-the-badge&labelColor=F05033&logo=git&logoColor=white)](https://github.com/thatrebeccarae/get-content-done)

<br>
<br>

```bash
git clone https://github.com/thatrebeccarae/get-content-done.git
```

<br>
<br>

<img src="assets/terminal-preview.png" alt="GCD terminal install preview" width="600">

<br>
<br>

[Why I Built This](#why-i-built-this) · [Who This Is For](#who-this-is-for) · [Getting Started](#getting-started) · [What It Does](#what-it-does) · [Skills](#skills) · [Agents](#agents) · [How It Works](#how-it-works) · [Extending GCD](#extending-gcd) · [Credits](#credits) · [Contributing](#contributing) · [License](#license)

</div>

---

## Why I Built This

Content production for solo builders is either chaotic or over-engineered. You're either dumping ideas into a Google Doc and hoping you remember to post, or wrangling a Notion database with 40 columns that nobody maintains. I wanted sprint methodology applied to content -- plan, produce, review, ship, retro -- with AI handling the parts it's good at (scoring, drafting, consistency checks) and humans handling the parts it's not (voice, judgment, "does this actually sound like me?").

## Who This Is For

- **Solo builders and creator-operators** who produce content across multiple platforms and need structure without a team
- **Content strategists** who want sprint methodology applied to content production, not just software
- **Claude Code users** who want a real-world skills + agents workflow they can extend and adapt
- **Obsidian and markdown-first writers** who refuse to move their content into another SaaS tool just to get a pipeline
- **Anyone publishing 3+ pieces per week** who needs quality gates, pillar enforcement, and performance feedback loops that actually close

## Getting Started

**Prerequisites:**

- [Claude Code](https://docs.anthropic.com/en/docs/claude-code) CLI installed and configured
- A markdown-based content store (Obsidian vault, flat files, whatever works)

**Install:**

```bash
git clone https://github.com/thatrebeccarae/get-content-done.git
cd get-content-done
./install.sh
```

**Configure:**

1. Copy `pillars.example.json` to your repo as `pillars.json`
2. Edit pillar names, posting days, keywords, and content types to match your strategy
3. Update vault paths in skill files to point to your content directories
4. Add signal briefs to your vault's inbox (manually via `/seed-idea` or from any source)
5. Run `/gcd:plan-sprint` to kick off your first sprint

<details>
<summary><strong>Path configuration</strong></summary>

Skills and agents reference three path categories that you'll need to update:

| Path | Default | What it points to |
|------|---------|-------------------|
| Repo path | `~/your-repo/` | Where you cloned GCD (for `pillars.json`, schemas, retros) |
| Content path | `~/your-vault/content/` | Where your content files live |
| Brief path | `~/your-vault/briefs/` | Where signal briefs are stored |

Search for these paths in the skill and agent `.md` files and replace with your actual paths.

</details>

## What It Does

GCD applies sprint methodology to content production. You define your content pillars, feed in signal briefs, and run a weekly cycle: **plan → produce → review → approve → retro**. Each step is a Claude Code slash command, and the planning step can also run autonomously via the morning pipeline (auto-assigning top briefs by impact score when no sprint-assigned stubs exist). Each piece of content tracks its own state in YAML frontmatter. Nothing moves directories. Nothing slips through the cracks.

**State layer** — `pillars.json` defines your content pillars, posting schedule, and quality gates. YAML frontmatter on every content file tracks lifecycle status. Files never move; status lives in frontmatter only.

**Scoring engine** — `analyze_content.py` scores every draft on a 5-category, 100-point scale before it reaches a human reviewer. Drafts that read like AI never make it past the auto-gate.

**Content templates** — 17 templates across LinkedIn, Substack, and adapted formats. Each template includes mandatory information gain markers (`[PERSONAL EXPERIENCE]`, `[ORIGINAL DATA]`, `[UNIQUE INSIGHT]`) that the producer must use instead of fabricating first-person content. Zero tolerance for invented anecdotes or fake metrics.

**Skills + agents** — 10 Claude Code skills orchestrate the workflow. 7 specialized agents handle scoring, drafting, review, research, and analytics. You interact through slash commands; agents do the heavy lifting behind each one.

## Skills

| Skill | Function |
|-------|----------|
| `/gcd:status` | Sprint dashboard — pieces, pillar coverage, velocity trends, bottleneck detection |
| `/gcd:queue` | Signal brief queue with composite scoring and pillar-fit suggestions |
| `/gcd:plan-sprint` | Interactive sprint planning with agent-scored recommendations and pillar enforcement |
| `/gcd:produce` | Route-dispatched content drafting (LinkedIn, Substack essay, Twitter/X thread, newsletter) |
| `/gcd:review` | Editorial review (Gate 1) — auto-gates on `analyze_content.py` scoring thresholds, then invokes agent for editorial judgment |
| `/gcd:approve` | Human approval (Gate 2) — only pieces with `reviewed` status can be approved |
| `/gcd:reject` | Reject a draft with feedback — preserves `.draft.md` for anti-pattern learning |
| `/gcd:retro` | Sprint close with pipeline metrics, engagement analysis, and performance feedback loop |
| `/seed-idea` | Capture a manual content idea as a brief stub |
| `/write-from-signal` | Convert a signal brief into a finished draft via a separate agent chain (research-analyst → content-marketer → editor-in-chief → social-amplifier) with user gates between stages |

## Agents

| Agent | Role |
|-------|------|
| `gcd-sprint-planner` | Scores briefs against pillars using composite formula (keyword 50% + pillar fit 30% + route fit 20%), detects strategy drift, applies performance boosts from historical data |
| `gcd-producer` | Template-aware drafting by route — selects from 17 content templates, enforces information gain markers, runs 5-point voice validation before save. Reads voice guide before every piece. |
| `gcd-reviewer` | Editorial review — invoked by `/gcd:review` after auto-gate passes. Receives `analyze_content.py` scorecard as context, focuses on subjective editorial judgment: voice consistency, structure, hook quality (A/B/C/D grading), SEO criteria, template adherence, marker validation. Returns pass/revise/escalate. Critique only — never rewrites. |
| `gcd-researcher` | Topic research for essay route — consumes RSS feeds and web sources, produces structured research briefs with data points, counter-arguments, and content angles |
| `gcd-amplifier` | Cross-platform distribution — generates LinkedIn posts, Twitter threads, email subject lines, and pull quotes from finished content |
| `gcd-metrics-analyst` | Pipeline analytics — throughput rates, gat

[truncated…]