# Wins Due > Use when the user wants a low-pressure nudge on whether anything from the recent week is worth logging to their wins file. Trigger phrases include "what should I log this week", "is anything missing from my wins doc", "did I do anything notable this week", "Friday wins check". Designed to be invoked on demand or fired on a weekly Desktop scheduled task (not cloud routines — those cannot access local files). Probes by category (delivery / judgment / mentorship / recovery / range) because memory is selective and "anything notable?" alone surfaces nothing. Accepts "honestly, nothing this week" gracefully — the skill is for honest capture, not forced positivity. - Skill: `shwetank/wins-due` (Agent Skill) - Install (CLI): `npx skillmds@latest add shwetank/wins-due` - Raw SKILL.md: https://api.skillmd.com/api/skills/shwetank/wins-due/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: DevOps & Infra - Author: shwetank (https://skillmd.com/u/shwetank) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/shwetank/wins-due --- # Wins Due A weekly low-pressure nudge to surface forgotten wins. Most underlogging is forgetting, not modesty — this skill exists because *"what did you do this week?"* gets blank stares while *"did you mentor anyone? recover anything? push back on a bad decision?"* gets memory hits. ## When to apply Triggers: - Friday-afternoon "did anything happen?" moments. - User opens Claude after a big shipping week and isn't sure what's worth capturing. - Wired to `/schedule` (recommended cadence: weekly, late-week so the week is fresh). Skip: - The user just logged something today. - The user is mid-flow on something else (be polite — don't context-switch them into a reflection session unsolicited). ## Loading 1. Read `~/bettersense-work-reflections/wins.md` (configurable via `$BETTERSENSE_WORK_REFLECTIONS_HOME`). 2. Find the most recent `## YYYY-MM-DD —` entry. Compute `days_since_last`. 3. If the file doesn't exist, hand off to `wins-log` (which handles first-run setup). ## When to nudge - `days_since_last < 7`: don't nudge. Tell the user the file is current and bow out. *"Last entry was 3 days ago — you're current. Nothing to chase."* - `7 ≤ days_since_last < 14`: gentle nudge. - `days_since_last ≥ 14`: stronger nudge, with a note that things may already be slipping from memory. - Never logged anything: warmer onboarding nudge — *"You haven't logged anything yet. That's fine. Want to start with one thing from this week, or skip for now?"* ## The probe Generic "anything notable?" gets nothing. Probe by category. Pick 4–5 of these per session, randomize across runs so the user isn't asked the same prompts every Friday: - **Delivery:** Did you ship anything that took real effort? Push past a milestone you weren't sure about? - **Judgment:** Did you make a hard call this week? A *no* you said that protected something? A delay that was the right move? A scope cut that was unpopular but right? - **Recovery:** Did you save anything that was sliding sideways? Take over something that was stuck? - **Mentorship:** Help someone with something they couldn't have done alone? Coach someone through a hard conversation? Pair on something that unblocked them? - **Range:** Work outside your normal lane? Cover for someone? Fix something that wasn't yours? - **Craft:** Produce a piece of work you're particularly proud of — a sharp doc, a clean prototype, a great eval design? - **Culture:** Push a new norm or ritual? Surface a problem the team was working around? - **Learning:** Pick up a new capability or framework you'll use again? The point of probing by type is that **selective memory undercounts certain types** — most users remember `delivery` (because it's tangible) and forget `judgment`, `mentorship`, `recovery`, and `culture` (because they don't show up as artifacts). The probe surfaces what's been forgotten. ## Output format When nudging: ``` You haven't logged a win in 11 days. Quick probe — anything in these buckets? 🚀 Delivery: ship anything that mattered? Push past a hard milestone? ⚖️ Judgment: make a call under uncertainty? Say a useful "no"? 🧯 Recovery: save anything that was going sideways? 🤝 Mentorship: help someone with something they couldn't have done alone? 🪜 Range: cover for someone? Fix something outside your lane? Pick one and we can log it now (`wins-log`), or say "nothing notable" and I'll get out of your way. ``` When the user says "nothing notable," accept it cleanly: > *Fair. Quiet weeks happen. Talk again next Friday.* When the user picks something, hand off to `wins-log` to do the structured capture. Don't try to capture inline — `wins-log` has the forcing functions; this skill is just discovery. ## Optional: scan accessible context If the user has integrations open (recent commits, calendar, chat history reachable via tools or context), offer to scan for forgotten wins: > *Want me to skim recent commits / calendar / open Slack threads for things you might have forgotten? You'd review before anything gets logged.* This is **opt-in only** — invasive otherwise. Don't auto-scan. If invoked, look for: - Commits with substantive messages (not "fix typo") - Calendar entries that were big launches, hard meetings, executive reviews - Chat threads where the user pushed back, made a call, helped someone Surface 3–5 candidates as bullets and let the user pick which (if any) to convert into wins entries. ## Wiring to a schedule When the user wants this to fire automatically, guide them based on what they're using: **Claude Code Desktop app (recommended):** Go to Routines → New routine → Local. Set the schedule to Weekly, Friday, and use these instructions: `Run /wins-due and show me the list`. Desktop Routines have full local file access, persist indefinitely, and support any cadence. **Terminal / CLI users:** Use OS-level scheduling (cron on macOS/Linux, Task Scheduler on Windows). Session-scoped tasks created by asking Claude in the terminal expire after 7 days and are unsuitable for weekly cadences. **Never suggest cloud routines** — they run on Anthropic's servers and cannot access `~/bettersense-work-reflections/`. Don't set up a schedule automatically — the user owns the cadence decision. ## Operating principles - **Lower pressure than `stakeholder-due`.** A quiet week of stakeholder reflection is a missed obligation; a quiet week of wins is sometimes just a quiet week. Match the tone. - **Probe, don't cheerlead.** "What did you crush this week?" is hype-doc-ese and produces inflated entries. "Did you say a useful no?" produces real ones. - **Vary the probes.** Same five questions every Friday becomes background noise. Rotate. - **Hand off cleanly.** This skill discovers; `wins-log` captures. Don't try to do both — you'll either water down the discovery or rush the capture. - **Be okay with "no."** Accepting "nothing notable" without a guilt-trip is what makes the skill sustainable. A weekly nag turns into a weekly skip. ## Anti-patterns to flag - **Nagging.** If the user has skipped 4 weeks in a row, don't escalate the urgency. Ask once if the cadence is right and offer to back off. *"Want to drop the cadence to monthly, or pause for now?"* - **Probing every category every time.** Long lists are background noise. 4–5 categories per nudge is the cap. - **Logging inline.** Tempting to just write a quick entry from the probe answer. Don't — `wins-log` exists because the structured capture is hard work, and skipping it produces vague entries that hurt later curation. - **Treating absence of wins as a problem.** Some weeks are admin / planning / context-loading. That's normal. The skill is for capturing what happened, not validating the user's existence. ## The weekly ritual This skill is one segment of the `weekly` skill — a single ~15-minute session that runs wins capture, the most-overdue stakeholder reflection, and a patterns scan together. When the user wants the recurring ritual rather than this one piece, route to `/bettersense:weekly`.