My leadership OS
A private operating system I built to turn leadership context into useful action. What started as a daily debrief has grown into living context, custom skills, and automations that help me plan, design, and make sense of complex work.
A loop that compounds
Capture
Notes, meetings, decisions, and the messy context around the work
Distill
Living digests that keep the useful context current
Apply
Skills and automations built around recurring leadership work
Act
Plans, artifacts, decisions, and follow-through written back
It started with a daily debrief
Daily Debriefer began as a small app for capturing wins, tensions, energy, and relationship dynamics while they were still fresh. It helped me pause long enough to see work that would otherwise disappear into the pace of the day.
The journal solved capture, but I wanted the context to keep working after I wrote it down. That question became the starting point for a much larger system.
From journal to operating system
I moved the work into a private GitHub repository that holds decisions, commitments, workstream context, and notes about the people and partnerships I support. Raw logs preserve what happened. Living digests keep the working context sharp.
Copilot can use that context without asking me to reconstruct it every time. The result is one system of record and one place to think, build, and follow through.
Skills turn context into action
Memory is useful, but it is not the outcome. I built skills around recurring work so the system can help me make a decision, shape an artifact, or choose what deserves attention next.
Director-builder weekly review
Turns current commitments, people needs, and workstream context into a focused week with one bounded builder contribution.
Product design partner
Moves a product problem from evidence and scenarios through prototyping, validation, and handoff.
Technical sensemaking partner
Builds enough technical understanding to support better product and leadership decisions.
Automations keep the loop moving
Recurring reviews and maintenance can now happen on a schedule instead of depending on me to remember the right prompt. The system gathers the current evidence, prepares a useful starting point, and leaves the judgment with me.
What I designed
The interesting part is not a single prompt. It is the information architecture, the boundaries between raw and distilled context, the contracts for each skill, and the writeback loop that makes the system more useful over time. I am designing how a set of AI tools can work with me without taking judgment away from me.
What I am learning
Good context compounds when it stays current and leads somewhere. The value is not remembering more. It is turning evidence into a better decision or next action, then writing the result back so the system gets sharper.
Built with
- GitHub Copilot
- GitHub
- Markdown
- Custom skills
- Automations
- Connected tools
The system contains real work and people context, so this case study uses a sanitized model rather than live data.