/repo-publisher

17 April 2026#reference#skill#orchestrator#publishing#repo-publisher

Stateful blog orchestrator that turns your git history into published blog posts. Reads your repo, analyses what changed, scores story candidates, writes drafts, and publishes through the scsiwyg MCP. Between runs, nothing is forgotten.

Trigger: "what should I blog about", "run the blog workflow", "initialize blog state for this repo", "draft a post about X", "publish the queue" Output: Published blog posts + updated .blog-state/ Type: Orchestrator (6-phase pipeline)

Workflow

Phase	What happens
1. Read State	Load `.blog-state/state.json`, compute git changeset since last run
2. Analyse	Run analysis pipelines (code-audit, security, infra-cost) on changed files
3. Decide	Score story candidates on significance × novelty × timeliness × audience-fit
4. Produce	Generate diagrams and tables, write post in markdown with excerpt and tags
5. Publish	Hand approved drafts to scsiwyg via `publish_post`
6. Save State	Snapshot state, update git baseline for next run

Decision Engine

Every story candidate is scored on four dimensions. Only stories clearing the configured threshold become posts.

Dimension	What it measures
significance	how much changed, how important is it
novelty	have we written about this before
timeliness	is this relevant now or will it go stale
audience_fit	does the target blog's readership care

A dependency bump is not a post. A major architecture refactor is. The threshold is configurable in config.yaml.

config.yaml

target_blog: your-scsiwyg-username
voice: "technical but accessible, first-person, no hype"
publish_mode: draft        # draft | auto | propose
cadence: weekly
pipelines:
  code_audit: on_change
  security: on_change
  infra_cost: on_dependency_change

publish_mode: draft queues posts for your review before anything goes live. auto publishes immediately. propose surfaces candidates without writing — you pick which ones to develop.

The .blog-state/ Directory

.blog-state/
├── state.json          # Project identity, git baseline, content ledger
├── config.yaml         # Target blog, voice, cadence, pipelines
├── history/            # Date-stamped state snapshots
├── queue/              # Drafts awaiting review
├── published/
│   └── log.jsonl       # Append-only publish record
├── media/              # Generated diagrams and charts
└── analysis/           # Cached pipeline outputs

This directory lives inside your repo and commits alongside your code.

Invocation

"initialize blog state for this repo"   → first-run setup
"what should I blog about"              → Phases 1–3, candidates only
"run the blog workflow"                 → All 6 phases end to end
"draft a post about the new auth system"→ Phase 4 with a specific topic
"publish the queue"                     → Phase 5 on existing drafts
"show blog state"                       → state.json summary

Skill Structure

repo-publisher/
├── SKILL.md
└── references/
    ├── config-schema.md
    ├── decision-engine.md
    ├── media-pipeline.md
    ├── state-schema.md
    └── story-templates.md

Use Cases

Download the full toolkit → · Back to library →

𝕏 Post