All skills

HR + Legal

Handle terminations, leaves, and investigations without the panic

The termination, the investigation, the accommodation request — the moments where HR and legal overlap and mistakes get expensive. Four free Agent Skills turn improvisation into a documented, consistent process. Learn the checklists and run them in Claude.

ClaudeChatGPTHR & People teamsManagersLegal
Flat editorial illustration of the employment lifecycle as a calm path — structured hiring, performance calibration, and a governed termination checklist under a lock

Why do employment actions go wrong — and how do you prevent it?

Employment actions carry real legal risk, and the failure mode is almost never bad intent — it’s inconsistency. The rules live in a compliance binder no one opens under pressure, so under pressure, everyone improvises. One manager documents; another doesn’t. A performance problem surfaces for the first time in the termination meeting because it was never written down.

Preventing it isn’t about knowing more law — it’s about running a consistent, documented process every time. That’s what these skills encode: the checklist, the risk gate, and the paper trail that turns the hardest conversation into a defensible one. And most of that defensibility is built long before the hard moment — in a clear role definition, a structured interview, a calibrated review.

The frameworks behind it

The skill isn’t magic — it encodes proven frameworks anyone can learn. Here’s the method itself, so you understand what it’s doing for you.

The documentation gate

employment-action-navigator

The single most important thing this skill does is stop you before you act. Before it will walk you toward a performance termination, it runs a documentation check: were expectations communicated in writing? Was a PIP completed? Was there a real opportunity to improve? If the answer is no, it says do not proceed — build documentation first.

This is the discipline that separates a defensible action from an expensive one. Most wrongful-termination exposure comes not from bad intent but from inconsistency — one manager documents, another doesn’t; one termination gets a risk check, the next gets fired off on a Friday. A gate the process runs every time removes that inconsistency.

Protected-class risk assessment

employment-action-navigator

Before any termination clears, the skill runs a risk assessment most managers wouldn’t think to run. It checks protected-class considerations (is the employee over 40, pregnant, on an ADA or religious accommodation, a member of a protected group?), recent protected activity (did they file a complaint, request FMLA, report a compliance concern in the last 6–12 months?), and timing concerns (does the action follow closely after protected activity? were the performance issues documented after it?).

If any box is checked, it routes you to legal review before you act — which is the entire point. The routine 80% moves fast and self-serves; the risky 20% reaches counsel before the mistake, not after. The skill also carries FMLA/ADA workflows, a five-step investigation framework, RIF/WARN-Act analysis, and state-specific final-pay tables.

The skills that get you there

Use any one on its own, or chain them — each hands its output to the next. Every one is free in the public library.

  1. 1

    Employment Action Navigator

    employment-action-navigator

    Process checklists and risk assessment for the hard actions — terminations, leaves, investigations — with a gate that routes protected-class or recent-protected-activity cases to legal first.

  2. 2

    Job Description Architect

    job-description-architect

    Builds leveled, competency-based, DE&I-optimized role definitions — the baseline every later performance and termination decision measures against.

  3. 3

    Interview Scorecard Designer

    interview-scorecard-designer

    Structured, behavior-anchored scorecards that turn “a bad feeling” into a documented, comparable, defensible hiring record.

  4. 4

    Performance Review Calibrator

    performance-review-calibrator

    Bias-checked, calibrated reviews that create the contemporaneous paper trail the navigator will demand before any action.

When you’d reach for it

Any HR or manager decision where a mistake is expensive and consistency matters — which is most of them. If your process for a termination or investigation depends on which manager is handling it, you have the inconsistency problem these skills solve.

  • Terminations — performance, conduct, or a reduction in force
  • Workplace investigations (harassment, discrimination, misconduct)
  • FMLA and ADA leave and accommodation requests
  • Building the hiring and review process so later decisions are defensible

What these skills do

There’s a category of work no one puts on a job description but everyone dreads: the termination, the harassment complaint, the accommodation request that arrives the week after a performance write-up. These are the moments where HR and legal overlap — and where “we’ll just handle it” quietly becomes a claim.

The navigator is the anchor — it turns those hard actions into risk-gated checklists. The People trio closes the gaps that seed most employment risk long before the hard conversation: a defensible role definition (job-description-architect), a structured hiring record (interview-scorecard-designer), and calibrated, documented reviews (performance-review-calibrator). Together they make the whole employment lifecycle defensible.

Try asking it

Real prompts, and what the skill hands back.

You ask

We want to terminate an employee for performance. Walk me through it — what documentation do I need, and what would make this risky?

What comes back

A decision tree with a documentation gate, then a protected-class/timing risk assessment that routes to legal review if any flag is present.

You ask

An employee filed a harassment complaint. Give me the investigation process and tell me when I must involve legal.

What comes back

The five-step investigation framework (assess → plan → investigate → conclude → follow up) plus the always-involve-legal triggers.

What makes the process hold up

A documentation gate

It won’t proceed without the paper trail — the check most managers skip under pressure.

Risk assessment first

Protected-class, recent-activity, and timing checks catch the exposure before you act.

Legal routing

The risky 20% reaches counsel first; the routine 80% self-serves. Effort matches risk.

A documented lifecycle

Structured hiring and calibrated reviews build the record long before the hard conversation.

Jurisdiction-ready

State tables and retention schedules are fill-in-the-blank — it becomes your compliant playbook.

How they chain across the lifecycle

Read them in order and they’re one lifecycle: define the role (job-description-architect) → hire on a structured record (interview-scorecard-designer) → manage with calibrated reviews (performance-review-calibrator) → act with risk-gated checklists (employment-action-navigator). Each stage builds the record the next depends on.

The governance kicker: the navigator’s termination checklist is exactly the skill you share use-only — a manager runs it without ever reading the firm’s risk thresholds or escalation logic. See use-only sharing.

Who it’s for

  • HR and People teams who own the hard, high-stakes decisions
  • Managers who need to act consistently and defensibly under pressure
  • Founders and COOs building a people process that scales without legal exposure

Do it yourself — run a defensible process

You don’t need employment counsel on retainer for every routine action. Run the navigator on the situation in front of you, and customize it with your jurisdiction’s rules:

Point Claude at the right skills

MCP server:agentman_skills
employment-action-navigatorjob-description-architectinterview-scorecard-designerperformance-review-calibrator

Already connected your library to Claude? Just tell it: “Use the employment-action-navigator skill (and the others below) from the agentman_skills MCP server.” Not connected yet? The Try in Claude / ChatGPT button above loads the first one from the public library — no setup. (connect your library.)

  1. 1

    Load the skill

    Click “Try in Claude / ChatGPT” above, or — if you’ve connected your library — tell Claude: “Use the employment-action-navigator skill from the agentman_skills MCP server.”

  2. 2

    Describe the action

    Tell it what you’re facing — a performance termination, an investigation, a leave request. It routes you into the right decision tree.

  3. 3

    Pass the documentation gate

    It checks whether the paper trail exists (written expectations, PIP, feedback). If not, it tells you to build documentation first — before you do anything you can’t undo.

  4. 4

    Run the risk assessment

    It walks the protected-class, recent-activity, and timing checks. Any red flag routes you to legal review before you act.

  5. 5

    Customize for your state

    Fill in the fill-in-the-blank state tables — final-pay timing, retention schedules, mini-WARN thresholds — so it becomes your playbook, not a generic one.

  6. 6

    Build the upstream record

    Use job-description-architect, interview-scorecard-designer, and performance-review-calibrator so future actions rest on a documented foundation, not memory.

Consistent process beats improvisation — and the difference is a defensible action instead of an expensive one.

Try Employment Action Navigator right now

One click loads the skill into Claude or ChatGPT — no account, no setup.

FAQ

Employment Action Navigator: common questions