Product ops skillsfor AI agents
31 open-source skills for the work that keeps product teams running: launches, customer feedback, metrics, planning, decisions, and docs. For Claude Code, the Claude app, Codex, Cursor, and any agent that reads skills.
One context file. Six kinds of work.
Run product-ops-context once. It writes down your product, teams, who decides what, your meetings, your tools, and your metric definitions. Every other skill reads that file first, so it only asks you for what's specific to the job.
What every skill refuses to do
The same four rules run through the whole library, so outputs hold up when someone checks them later.
Every open item gets one named owner.
"Marketing" isn't an owner. Neither is "the team."
Discussed and decided stay separate.
A lean in a meeting never shows up as a decision.
Nothing gets invented.
If the notes don't say why, the output says "Not recorded."
Skills recommend. People decide.
The agent does the sorting and drafting. A person makes the call.