Employment law doesn't pause for AI adoption.
HR teams are deploying agents for candidate screening, performance reviews, compensation analysis, and onboarding. The skills driving those agents were written by HR analysts who understand the workflow — but who may not hold the full legal and policy review those processes require.
HR processes carry legal obligations. Agent skills don't inherit them automatically.
The recruiting team deployed an agent to handle initial candidate screening. The skill was written by a senior recruiter who had been doing the job for six years. She knew what to look for. She also, without realizing it, encoded several criteria that employment counsel would flag immediately — proxies for protected characteristics, questions that aren't legally permitted in certain jurisdictions, a weighting methodology that hadn't been reviewed for disparate impact.
The skill passed no review. There was no review process. It went live on a Tuesday and was running at full capacity by Thursday. The agent screened 400 candidates in its first two weeks.
Nobody did anything wrong in the usual sense. The recruiter applied the best judgment she had. The skill worked — response rates were up, time-to-phone-screen was down, the hiring manager was happy. The exposure wasn't visible until someone asked: who reviewed the screening criteria?
The answer was nobody. Because there was no system for reviewing agent skills. Because HR never had to review skills before — they reviewed people's judgment. The judgment is now running in software, at scale, and the review process hasn't been redesigned to match.
Institutional HR knowledge walks out the door when people leave.
Every HR team has people who carry institutional knowledge that isn't written down. How the performance review process actually works versus how it's documented. The unwritten rules about how promotions get calibrated. The way that certain types of performance issues get handled in practice. The senior HRBP who has been managing the process for ten years and holds all of this in her head.
When those people leave — and they leave — the knowledge leaves too. Sometimes it ends up in a Confluence doc that goes stale. Sometimes it doesn't end up anywhere. The people who come after have to rediscover it through mistakes or ask the wrong questions.
Agent skills that encode institutional HR knowledge — done right, with review — are how that knowledge becomes durable. The performance calibration methodology becomes a skill that the next generation of HRBPs runs from day one. The onboarding process standard becomes a skill that the agent follows whether the onboarding manager is new or has been there for eight years.
Review processes that match the risk.
Legal review for high-risk processes
Candidate screening, performance evaluation, compensation analysis, and termination workflows carry legal exposure. Skills encoding those processes need employment counsel in the approval chain — not as an afterthought, as a required gate before the skill reaches production.
Policy alignment
HR skills should reflect your current policy, not the policy that was in place when the skill was written. When your PTO policy changes, the onboarding skill changes with it. When your performance review methodology is updated, every agent running performance workflows picks up the new version.
Captured expertise
The judgment of your best HRBPs, your most experienced recruiters, your most thorough compensation analysts — captured as skills, versioned, and deployed as the default behavior for the agents running those processes. Institutional knowledge that doesn't walk out the door.
What Invoked does for HR governance.
Invoked provides the governance layer between HR process design and agent execution — specifically the review and approval infrastructure that HR skills require.
Authoring
HR professionals, HRBPs, and process owners author skills through a structured workflow that captures the full process logic: standard steps, edge cases, escalation triggers, policy requirements. The authoring path enforces completeness — every skill has an owner, a declared scope, and the policy reference it implements.
Tiered review
Low-risk HR skills (onboarding logistics, FAQ responses, scheduling) move through a lightweight review. High-risk skills (screening criteria, performance evaluation methodology, compensation analysis) require employment counsel and HR leadership sign-off. The review tier is determined by the skill's declared scope — not by whoever happens to notice.
Audit trail
Every skill invocation in a people process is logged with the full context: which skill version ran, who approved it, what logic was applied, what the outcome was. When an employment claim is filed, the records exist. When a regulator asks whether the screening process was reviewed, the answer is yes — and here is the approval chain.
Understand what your HR agents are running.
Before you can govern HR agent skills, you need to see what exists. Invoked reads the skill paths your agents discover from — read-only. You get a map of every skill in your HR stack: what process it governs, who built it, when it was last reviewed, whether it has a legal sign-off on record. Most CHROs find skills running processes that were never reviewed by employment counsel. Some find skills that encode criteria that would concern a labor attorney. A few find skills that were built for a process that has since been updated — but the agent hasn't been.
What comes after.
If what we find is material, we run a 90-day paid pilot with one HR function — typically recruiting or performance, whichever carries the most agent volume. Skills authoring. Tiered review workflow with employment counsel integration. Fleet deployment. By the end, your HR agents are running reviewed, legally-signed-off skills that match your current policy — not informal ones built without oversight.