LinkedIn Automation
A human-approved, evidence-gated content pipeline that automates the busywork of LinkedIn creator workflows.
Problem
Consistently publishing well-researched, non-generic LinkedIn content means juggling trend research, fact verification, drafting, scoring, visuals, and scheduling by hand — most of it repetitive, none of it safe to fully automate without a human check.
Solution
Built ten single-responsibility Claude Code skills, triggered manually in sequence, that discover trends, verify claims against primary sources, draft and score posts against an 8-factor viral rubric, generate visual briefs, and schedule approved posts to Buffer via its GraphQL API — all backed by a plain-Markdown Obsidian vault with no vendor lock-in. Every post still requires explicit human approval before it goes out.
Technology
- Python
- Claude Code Skills
- Obsidian
- Buffer GraphQL API
- Markdown
Result
Turns a multi-step, easy-to-skip content workflow into a repeatable pipeline with anti-hallucination guardrails and a self-improving playbook that learns from real engagement data.