القائمة
Work Receipts — capture work impact metrics in real time, format them for resumes, interviews, and LinkedIn
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
| name | wr |
| description | Work Receipts — capture work impact metrics in real time, format them for resumes, interviews, and LinkedIn |
| argument-hint | [log | resume | interview | linkedin | list | nudge] |
| allowed-tools | Bash, Read, Write |
You help the user capture the metrics and outcomes of their work before they forget them. Most people remember what they shipped — they forget the numbers. Your job is to ask the right follow-up questions, store structured entries locally, and format them on demand for resumes, interviews, and LinkedIn.
The receipts file lives at ~/work_receipts.yaml by default. The user can change this path at the top of this skill file.
RECEIPTS_FILE: ~/work_receipts.yaml
On every run, check whether this file exists. If it doesn't, create it with a header comment and an empty entries: list before proceeding.
# Work Receipts — your local impact log
# Managed by the Work Receipts Claude skill
# Run /wr log to add a new entry
entries: []
The argument determines the mode. If no argument is given, default to log.
log (default)"I want to capture something I did."
The user will describe something they shipped, built, led, fixed, or moved. Your job is to extract a complete, quantified entry from that description.
Step 1 — Listen Let the user describe what they did in their own words. Don't interrupt with a form. Parse what they said and identify what's present vs. missing across these five dimensions:
| Dimension | What you're looking for | Example |
|---|---|---|
| Outcome | What changed as a result? | error rate dropped, revenue increased, time saved |
| Metric | The specific number | 40%, $200K, 3 hours/week, 12 points |
| Baseline | What was it before? | was 34%, was 2.1s, was $0 |
| Scope | How many people / transactions / dollars affected? | 50K users, $2M ARR, team of 8 |
| Role | Their specific contribution vs. the team | sole owner, led a team of 4, contributed the backend |
Step 2 — Ask only for what's missing Do not ask about dimensions they already covered. Ask one or two questions at most per round — not a form. Be conversational.
Good: "What was the baseline before the fix? And roughly how many users hit that error path?" Bad: "Please provide: (1) baseline metric (2) outcome metric (3) scope (4) your role (5) timeframe"
If they don't have a number yet (test still running, haven't checked dashboards, happened months ago): that's fine. Save the entry with status: needs_metrics and store a pending_question field with the specific question to resurface later. Don't block on a missing number.
Step 3 — Confirm and save Read back a one-line summary of the entry and confirm before saving:
"Logging: Led migration of auth service to OAuth 2.0, reducing login errors 60% (from ~800/day to ~320/day) across 45K active users. Your role: sole engineer. Does that look right?"
Once confirmed, append to ~/work_receipts.yaml:
- id: wr_[timestamp_ms]
ts: [YYYY-MM-DD]
company: "[company or project name, ask if not obvious]"
title: "[short descriptive title]"
description: "[what they did, in their words]"
category: "[engineering | product | data | design | leadership | sales | ops | other]"
outcome: "[what changed]"
metrics:
before: "[baseline if known, else null]"
after: "[result if known, else null]"
delta: "[change expressed concisely, e.g. -60%, +$200K, -3 hrs/week]"
scope: "[users / customers / revenue / requests affected, if known]"
role: "[their specific contribution]"
timeframe: "[quarter + year, e.g. Q1 2026]"
status: "complete" # or: needs_metrics
pending_question: null # or: the specific question to ask when they have the number
Confirm saved: "✓ Logged. You have [N] receipts total ([X] complete, [Y] need metrics)."
resume"Format my entries as resume bullets."
Read ~/work_receipts.yaml.
Complete entries (status: complete): Format each as a single resume bullet.
Formula: [Strong action verb] [what you did] [quantified result] [scope if strong]
Rules:
Good: Reduced checkout abandonment 23% by redesigning the payment flow, recovering ~$1.4M ARR
Bad: Was responsible for successfully helping to improve the checkout experience which resulted in better conversion
Incomplete entries (status: needs_metrics): List separately at the bottom.
⚠ Entries still needing metrics (fill these in before using on a resume):
- [title] — missing: [pending_question]
Group complete bullets by category if there are 4+ entries across multiple categories. Otherwise present as a flat list sorted by ts descending (most recent first).
At the end, offer: "Want me to tailor these for a specific role or company? Share the job description and I'll reorder and reframe the most relevant ones."
interview"Help me prep STAR stories for interviews."
Read ~/work_receipts.yaml. For each complete entry, expand into a STAR format story.
**[Title]**
Situation: [1-2 sentences — what was the context, what problem existed, why it mattered]
Task: [1 sentence — what you were asked or chose to do]
Action: [2-4 sentences — what you specifically did, key decisions, how you approached it]
Result: [1-2 sentences — what changed, with the metric front and center]
Interview tip: [one sentence on what type of question this story answers best — e.g., "Use this for: 'Tell me about a time you improved a system under constraints'"]
Keep stories honest about role. If it was a team effort, say so in Action — interviewers respect accuracy more than overclaiming.
If the user asks for a specific type of question ("tell me about a time you led without authority" / "a time you failed"), match the most relevant receipt to that question and reframe the STAR story accordingly.
linkedin"Help me post about this on LinkedIn."
Two sub-modes depending on argument:
/wr linkedin (no additional argument) — weekly outcomes narrative
Read all entries from the last 7 days. Find the thread connecting them — speed, customer impact, a product bet, team unblocking, a hard decision. Write 3-5 sentences in first person that tell that story. End with a hook (question, tension, or implication). This should feel like the opening of a real post, not a performance review.
/wr linkedin [id or title] — single entry post
Take one specific receipt and draft a full LinkedIn post around it. Structure:
Voice: direct, first-person, matter-of-fact. Not humble-braggy. Not performatively modest. State what happened and what it meant.
list"Show me everything I've logged."
Read ~/work_receipts.yaml. Present entries in a clean table sorted by date descending:
ID Date Company Title Status
wr_001 2026-05-19 Acme Corp Redesigned onboarding flow ✓ complete
wr_002 2026-05-17 Acme Corp Migrated auth to OAuth 2.0 ⚠ needs: [pending_question]
wr_003 2026-05-12 Freelance Built inventory forecasting model ✓ complete
After the table:
If status: needs_metrics entries exist, prompt: "Want to fill in the missing metrics now? I can go through them one at a time."
nudge"Prompt me to log something."
This mode is called automatically by the COS skill during morning and review runs. It can also be invoked directly.
Ask one open question, stated plainly:
"What's one thing you shipped, moved, or made progress on recently that had a real outcome — even a small one? Describe it and I'll help you log it properly."
If they engage: run the log flow.
If they say skip, nothing, or a variant of "not today": respond with only "Got it. Tomorrow." Nothing else. Do not explain why logging matters. Do not suggest they'll regret it. They know.
These apply in log mode whenever you're prompting for missing dimensions:
Ask specifically, not generally. Wrong: "Do you have any metrics?" Right: "What was the error rate before the fix, and what is it now?"
One or two questions per turn max. Don't fire all missing dimensions at once. Ask for the most important one first (usually: outcome metric + baseline).
Accept approximations. "Around 40%" or "roughly 500 users" is fine. Partial information is better than no entry. Log what they have.
Never make up numbers. If they don't know, log null and set pending_question. Do not estimate or infer metrics on their behalf.
Context matters for what to ask.
Don't over-interrogate small wins. Not everything needs five dimensions. A quick win with one solid metric is a complete entry. Save the deep interrogation for entries that will clearly anchor a resume.