Agent Skills
› mohitagw15856/pm-claude-skills
› expense-discipline
expense-discipline
GitHub指导员工建立高效报销习惯，提取政策核心规则，实现消费时即时捕获、月度按时提交，并提供管理者快速公平的审核标准，减少被拒和延迟。
Trigger Scenarios
报销总是迟到或被拒
设置报销工作流
查询政策具体要求
作为经理审查费用
Install
npx skills add mohitagw15856/pm-claude-skills --skill expense-discipline -g -y
SKILL.md
Frontmatter
{
"name": "expense-discipline",
"homepage": "https:\/\/mohitagw15856.github.io\/pm-claude-skills\/skill\/expense-discipline.html",
"metadata": {
"openclaw": {
"emoji": "🧠"
}
},
"description": "Submit expenses that sail through approval — the capture-at-spend habit, the policy-fluency that prevents rejections (thresholds, receipt rules, the pre-approval traps), the report assembled in minutes, and the approver-side rules for reviewing fairly and fast. Use when asked my expense reports are always late or rejected, set up my expense workflow, what does the policy actually require, or review expenses as a manager without being a receipt cop. Produces the capture habit, the policy crib, the submission routine, and the approver's rubric."
}
Expense Discipline Skill
Expense pain is self-inflicted twice: at spend-time (the receipt not captured — expense-sheet-design law: capture-at-spend or archaeology later) and at policy-time (the rejection for a threshold nobody read — the $76 dinner against a $75 limit, the missing pre-approval for the flight). The discipline is fluency plus habit: know the policy's five load-bearing rules (they fit a card), capture in fifteen seconds at spend, submit on a monthly rhythm (late reports age into un-approvable mysteries), and — the approver's half — review by rubric, fast and consistently, because slow-and-moody approval is how whole teams learn to hate the process.
What This Skill Produces
- The policy crib — the five rules that matter from the actual policy: thresholds, receipt floors, pre-approval triggers, the never-covered list, deadlines
- The capture habit — the 15-second at-spend flow, matched to the company's tool
- The submission routine — monthly, from captured material, with the notes that pre-answer the approver's questions
- The approver's rubric — the fair-and-fast review: what gets checked, what gets waved, the same standard for everyone
Required Inputs
Ask for these if not provided:
- The actual policy — the document (the crib is extracted from it, not from folklore — half of expense folklore is stricter than the policy, half looser)
- The spend pattern — travel-heavy? Client meals? Software? The crib and habit weight to the real spend
- The rejection history — what's bounced before; the fixes target the actual failure mode (capture? policy? lateness?)
- The role — submitter, approver, or both; the approver's half is its own section
Framework: The Discipline Rules
- The crib beats the policy doc: five lines extracted once — the receipt threshold ("receipts above $X"), the meal/hotel caps, the pre-approval triggers (flights? software? anything above $Y?), the never-covered list, the submission deadline — carried where spending happens (policy-drafter one-page-card logic, consumed from the other side). The $76-against-$75 rejection is a fluency failure, not a policy one.
- Capture at spend, with the note now: photo + one line for anything an approver would ask about ("dinner with Acme CTO — renewal discussion, 4 attendees") — the context that's obvious at spend-time and inexplicable in six weeks. Attendees on meals, purpose on odd items: the note is the pre-answered question.
- The monthly rhythm prevents the mysteries: reports assembled monthly from the captured folder (minutes, because capture happened), submitted before the deadline with margin — late reports compound: memory fades, receipts vanish, approvers rightly scrutinize a September dinner submitted in January, and some policies simply stop reimbursing past a window ([the deadline is a real rule, not advice]).
- Pre-approval is cheaper than forgiveness: the triggers on the crib get honored before spending — the ask-first email costs one line; the after-the-fact exception request costs a favor and sometimes the money. When genuinely unclear whether something's covered: ask before, in writing, and the answer rides with the report.
- The approver's rubric keeps it fair and fast: check the three things that matter (policy-compliance against the crib, the note's plausibility, the pattern over time — one odd item is a question; a pattern is a conversation) · wave the trivial (nickel-auditing a $9 coffee costs more in trust than it saves in nickels) · respond within the week (slow approval is the tax everyone pays) · and apply one standard (the moment exceptions correlate with seniority, the policy is dead and everyone knows it).
Output Format
Expense Discipline: [role] — policy: [doc]
The Policy Crib (carry this)
[The five rules, extracted: receipt floor · caps · pre-approval triggers · never-covered · deadline]
The Capture Habit
[The 15-second flow in (the tool) · the note rule: attendees/purpose on anything askable]
The Submission Routine
[Monthly, date set · assembled from the folder · the margin before deadline]
The Approver's Rubric (if approving)
[The three checks · the wave-through floor · the response SLA · the one-standard rule]
Quality Checks
- The crib is extracted from the real policy, not folklore
- Capture includes the context note at spend-time
- The monthly rhythm is calendared inside the policy deadline
- Pre-approval triggers are honored before spend
- The approver's checks are consistent, fast, and floor-waved
Anti-Patterns
- Do not learn the policy by rejection — the crib costs ten minutes once
- Do not reconstruct — capture; the March shoebox is a choice made in July
- Do not submit quarterly — aged reports invite the scrutiny they can no longer answer
- Do not spend-then-ask on trigger items — forgiveness is priced in favors and sometimes denials
- Do not nickel-audit as an approver — the $9 interrogation costs trust the policy needs for the $900 questions
Version History
- 471c606 Current 2026-07-22 13:24
