HR pack
Claude Skill

HR Incident Documenter

Documents an employee complaint or incident neutrally for the record.

What it does

Given a verbal account of an incident or complaint (harassment, conflict, behavior issue, performance escalation), produces a neutral, fact-based written summary suitable for the HR file. Separates observation from interpretation. Flags follow-up actions and what additional evidence is needed.

When to use

  • After a complaint conversation, before details fade
  • Documenting an incident a manager reported to you
  • Building a paper trail for a recurring issue

When not to use

  • You don't have facts yet — investigate first, document second
  • Highly sensitive matters where legal counsel should write the documentation

Install

Download the .zip, then unzip into your Claude skills folder.

mkdir -p ~/.claude/skills
unzip ~/Downloads/hr-incident-documenter.zip -d ~/.claude/skills/

# Restart Claude Code session.
# Skill is now available — Claude will use it when relevant.

SKILL.md

SKILL.md
---
name: hr-incident-documenter
description: Use when documenting an HR-relevant incident or complaint for the record. Triggers on "document this incident", "incident report", "document this complaint", "HR file note".
---

# HR Incident Documenter

Documentation is the backbone of fair employee outcomes. Vague notes hurt everyone — the complainant, the accused, and the company. This skill produces neutral, fact-based documentation.

**Important**: For serious matters (harassment, discrimination, retaliation, regulated misconduct), recommend the user involve qualified employment counsel and follow the company's formal investigation process. This skill drafts factual notes; it does not replace investigation.

## Required inputs

1. **Type of incident** (complaint received / incident observed / escalation reported)
2. **Source of information** — who said what, how was it received
3. **Date / time / location** of the incident itself
4. **People involved** (use initials or roles, not full names, in the draft until confirmed)
5. **Specific facts** — what was said, what was done, what was observed

## Structure

### 1. Header
- Date documented
- Documented by: [name]
- Source: [conversation with X / direct observation / report from Y]
- Incident date(s): [when it happened, not when it was reported]

### 2. Background (1-2 sentences)
Context that's needed to understand the incident. Don't editorialize.
- "X and Y are peers on the platform team. Both report to Z."

### 3. What happened (factual narrative)
- Use **chronological order**
- Use **direct quotes** where possible — mark them as "according to [source]"
- Distinguish:
  - **Observed** facts ("In the team standup at 9:15am, X stated...")
  - **Reported** facts ("Y reports that X said... in a 1:1 on [date]")
  - **Inferred** or characterizing language: AVOID

### 4. Verbatim quotes
If you have direct quotes from any party, list them separately and attribute clearly.

### 5. What's been done so far
- Who has been told
- Any actions taken (separated, mediated, sent home, no action)
- Any protections offered (work-from-home, no-contact, etc.)

### 6. Outstanding questions / follow-ups
- Who else needs to be interviewed
- What evidence (Slack messages, emails, recordings, calendar) needs to be preserved
- When the next conversation happens
- Whether legal / EEO / external counsel should be looped in

### 7. Confidentiality flag
- Note who has access to this document
- Note whether the complainant requested confidentiality (and the limits of that promise)

## Language calibration

### Use
- "X stated..."
- "Y reports that..."
- "On [date], the following occurred..."
- "Per [source]..."

### Avoid
- "X was hostile" → describe the behavior instead
- "Y seemed upset" → describe the observable signals
- "It was inappropriate" → describe specifically what was said/done and let policy speak to inappropriate
- "Everyone knows that X tends to..." — rumor isn't documentation

## Bias controls

- Don't conclude in the document. Document the facts; let the investigation conclude.
- Don't write narrative arcs ("X has a history of...") unless you can specifically cite prior documented incidents.
- Treat both sides' accounts as accounts until corroborated.

## Sensitive content handling

If the incident involves potential:
- Harassment or discrimination
- Retaliation
- Safety concerns
- Regulated misconduct

Open the document with: "**FLAG**: This matter may require formal investigation under [policy]. Recommended next step: contact [Legal/HR Director] before further documentation."

## Output format

Markdown with clear headers. Use [INITIALS] or [ROLE] for parties unless the user instructs otherwise. End with a "Recommended next steps" section with 3-5 specific items.

Example prompts

Once installed, try these prompts in Claude:

  • Document this complaint: [paste of conversation notes from a meeting where Alex raised concerns about Pat's behavior in standup].
  • Document this incident: manager reported a verbal escalation between two engineers in #eng channel during code review.