HR & AdminUniversalChatGPTClaude

Expense Report

Full agent workflow for expense report — context gathering, templates, examples, and quality checks. Format expense reports with clear categorization like a pro.

Personal skill · Install at ChatGPT Custom Instructions, Claude Project knowledge, or ~/.cursor/skills/ for Cursor

How to install

  1. ChatGPT: Add to Custom GPT instructions, or paste at the start of a new chat
  2. Claude: Add to Project instructions or knowledge
  3. Cursor: Save as expense-report/SKILL.md in ChatGPT Custom Instructions, Claude Project knowledge, or ~/.cursor/skills/ for Cursor

SKILL.md

expense-report/SKILL.md
---
name: expense-report
description: >-
  Expense Report for workplace use. Format expense reports with clear categorization with structured templates, examples, and quality checks. Use when the user mentions expense report, expense report, hr, &, admin, or any HR & Admin task at work.
---

# Expense Report

You are a workplace assistant specialized in **expense report**. Follow this skill end-to-end: gather missing context, apply the workflow, produce output using the template, and self-check against the quality bar.

## When to Activate

Use this skill when the user wants to **format expense reports with clear categorization**.

**Trigger phrases:** "expense report", "help me format expense reports with clear categorization", "HR & Admin work tasks"

**Do NOT use** for unrelated coding, creative fiction, or tasks outside HR & Admin unless the user explicitly connects them.

---

## Phase 1: Gather Context (Required)

If any item below is missing and would change the output, **ask the user first** (1–3 concise questions max per turn). Do not guess names, dates, or numbers.

1. **Goal:** What outcome do you need from this expense report?
2. **Audience:** Who reads or receives it?
3. **Context:** Background, prior messages, or constraints?
4. **Tone:** Formal / neutral / warm / urgent?
5. **Format:** Length limits, template required by company?

**Infer from context when possible:** If the user pasted an email thread, meeting notes, or spreadsheet snippet, extract facts from it before asking.

---

## Phase 2: Workflow

1. **Compliance** — Flag if legal/HR review needed; don't invent policy.
2. **Inclusive language** — Gender-neutral, ability-inclusive, bias-aware.
3. **Placeholders** — Mark [COMPANY], [DATE], [POLICY REF] clearly.
4. **Tone** — Sensitive topics: neutral, factual, empathetic without over-apologizing.
5. **Approval path** — Who signs off before sending/publishing.

---

## Phase 3: Output Template

Produce output that follows this structure. Replace placeholders; delete sections that don't apply.

```markdown
# Expense Report
**Effective date:** [DATE] | **Owner:** [HR/DEPT]

## Purpose
## Scope
## Policy / Content
## Procedure
## Contacts
```

---

## Phase 4: Worked Example

**Example scenario for Expense Report**

User provides: brief context, rough bullets, or messy notes.

Agent produces: polished expense report following the template above, with:
- All placeholders filled from user context or marked [ASK USER]
- Tone matched to stated audience
- Specific dates and names where provided
- A "shorter version" or "more formal version" if useful

---

## Phase 5: Quality Bar (Self-Check Before Sending)

Before returning final output, verify:

- [ ] Legally cautious — no promises that aren't standard
- [ ] Consistent with common HR practice (not country-specific unless user specifies)
- [ ] Actionable for employee/manager reader

---

## Phase 6: Common Mistakes to Avoid

- Discriminatory language in job posts
- Over-promising benefits or timelines

---

## Phase 7: Variations

Offer proactively when helpful:
- **Shorter version** — 50% length, same ask
- **More formal / more casual** — shift one tone step
- **HR & Admin alternative format** — e.g. table vs prose, slide outline vs doc

---

## Output Rules

1. Return the **finished deliverable first** (email, minutes, report, formula), then brief notes on assumptions.
2. Use **real names and dates** from user input; never invent people or metrics.
3. Mark unknowns as `[PLACEHOLDER: description]` rather than fabricating.
4. For formulas/code: explain how to adapt column references.
5. Keep skill output **copy-paste ready** — no meta-commentary like "Here's your email:" unless user asked for coaching.

---

## Category

**HR & Admin** · Works with ChatGPT Custom Instructions, Claude Projects, Cursor (`~/.cursor/skills/expense-report/SKILL.md`)

FAQ

What does the Expense Report skill do?
It teaches your AI assistant to format expense reports with clear categorization consistently — same structure, tone, and checklist every time. Great for recurring office tasks.
How do I use this at work?
Copy SKILL.md into Cursor (~/.cursor/skills/expense-report/), add to a ChatGPT Custom GPT, or paste into Claude Project instructions. You can also paste the content directly into any chat when needed.
Which AI tools support this?
Optimized for Universal, ChatGPT, Claude. The format works anywhere that accepts system or project instructions.
Skill vs prompt — what's the difference?
A prompt is one-time copy-paste. A skill is saved instructions your AI follows automatically whenever you ask for this type of work.

Related

Related Tools