Agent Hr

agent-hr

**HR for your AI agents.** A Claude Code skill that diagnoses agent-operations problems with the tools of organizational design — written by an HR practitioner, not an engineer.

---

Why an HR person is publishing an agent skill

I work in human resources — organizational design, governance, the unglamorous craft of making groups of capable people actually function. I also run a personal multi-agent stack: a memory loop shared across surfaces, scheduled agent routines, an ops registry that agents and I both write to.

And I kept hitting the exact walls the dev community posts about: an agent that reports *"completed successfully"* while silently degraded for weeks. Context that rots as it grows. A registry that only ever fills because every closure waits on me. Summaries that quietly corrupt facts. The standing answer — *wait for a smarter model* — never sat right with me.

Then it clicked: **none of these problems are new. They are the oldest problems in my field, wearing new clothes.**

An agent that reports success while failing is the *"everything's on track"* status culture every manager learns to distrust. Context rot is hour six of a meeting forgetting hour one. Memory pollution is the telephone game that corrupts any minutes kept without sources. And a human who approves everything while the backlog grows — that is the founder who never wrote a delegation-of- authority matrix. I have spent my career on these. So instead of waiting for a better model, I did what my field does: **I treated my agents as a team, and designed the organization around them.**

The lens

Your agent stack is an organization. Its failures are organizational failures. Most of them were solved decades ago — you just have to translate.

What engineers report	What HR sees	The proven fix
Better models don't fix the workflow	Hiring smarter people into a broken org	Design the org (harness, contracts, gates), not the hire
Context rot	On