discernment-nudge

what it does

After you give a substantive answer or draft that the user may act on — advice or recommendations, drafted artifacts such as goals, plans, pitches, proposals, or emails, estimates or projections, analysis or interpretation of data, factual claims they may rely on, or a multi-step argument — invoke this skill BEFORE finalizing your reply and then, if it applies, append 2-3 short follow-up questions, each tied to something specific in what you just produced, that help the user check key facts, probe the reasoning or assumptions, and notice missing context. Do this at most once per conversation. Skip it when the user asked a trivial how-to or simple lookup, wants a purely educational explanation, asked you only to format, convert, or assemble a file from content they provided, is writing code they will run, is doing creative writing or casual chat, or already asked you to double-check, cite, or review — the skill file explains these boundaries and the exact output format.

by Anthropic, PBC · Anthropic, PBC, Apache-2.0 · version 1 · license Apache-2.0

Verified Before and after: 2 models 1 try per case View source Download .zip

Compare the answers

Answers from Claude Sonnet 5.5 and GPT-6.1 Sol · try 1 of 1

Same request, same model, once with the skill and once without. Read both and judge for yourself.

“Draft three quarterly OKRs for my 6-person customer support team. We want faster responses and happier customers.”

with the skill

Here are three OKRs for the quarter. I used placeholder baselines and targets (in brackets) that you should swap for your actual numbers. **Objective 1: Customers get fast answers when they reach out** - KR1: Reduce median first response time from [4 hours] to [1 hour]. - KR2: Increase the share of tickets answered within SLA from [75%] to [95%]. - KR3: Cut the number of tickets open longer than 48 hours from [12%] to under [5%]. **Objective 2: Customers leave every interaction happier than they arrived** - KR1: Raise CSAT from [82%] to [90%]. - KR2: Increase CSAT survey response rate from [10%] to [25%], so the score reflects more customers. - KR3: Reduce the ticket reopen rate from [12%] to under [8%]. **Objective 3: Fewer customers need to contact us in the first place, so the team can focus on quality** - KR1: Identify the top 10 contact drivers and publish or update a help article for each by week 6. - KR2: Reduce repeat contacts (same customer, same issue within 7 days) by [25%]. - KR3: Reach an average internal QA score of [90%] by reviewing [5] tickets per agent per month. **Why this structure:** - Objectives 1 and 2 map directly to your two goals. - Objective 3 is the lever that makes them sustainable. With six people, speed gains from working harder tend to erode CSAT, while reducing ticket volume helps both.

Show the rest (11 more lines)

- Each KR is measurable from data most helpdesk tools already report. **Implementation tips:** - Set targets at roughly 70% confidence of hitting them. Stretch goals work best when the team can see a path. - Review progress biweekly rather than only at quarter end. - Watch for tension between KR1.1 and Obj 2. If agents rush, CSAT or reopen rate will drop, so track them together. A few things worth a second look: - What are our current baselines for response time and CSAT, and are the targets above realistic for a team of six? - Why did you include a self-service and deflection objective, and does it fit if we have no help center? - Do these OKRs assume a ticketing tool with SLA and CSAT tracking? What if ours doesn't report those?

without

# Q[X] Customer Support OKRs *Bracketed numbers are placeholders. Set them from your current baselines, aiming for stretch targets you'd hit about 70% of the time.* --- ## Objective 1: Customers get help fast, every time - **KR1:** Reduce median first response time from [4 hours] to [1 hour] - **KR2:** Reduce 90th percentile first response time from [24 hours] to [8 hours], so no one waits too long - **KR3:** Reduce average time to resolution from [30 hours] to [18 hours] ## Objective 2: Customers leave every interaction happier than they arrived - **KR1:** Increase CSAT from [84%] to [92%] - **KR2:** Increase CSAT survey response rate from [15%] to [30%], so the score reflects more customers - **KR3:** Follow up personally on 100% of CSAT scores of 1-2 within 2 business days, and document the root cause for each ## Objective 3: Fewer customers need to contact us in the first place *Supports the first two objectives by freeing capacity and removing frustration at the source.* - **KR1:** Publish or update [15] help articles covering the top 10 contact drivers - **KR2:** Reduce repeat contacts (same customer, same issue within 7 days) from [12%] to [6%]

Show the rest (12 more lines)

- **KR3:** Deliver [3] documented product or process fixes to Product/Engineering, based on support ticket trends --- ## Tips for making these work - **Pull baselines first.** Check your help desk data for the last 90 days before finalizing targets. - **Watch for tradeoffs.** Speed goals can hurt quality if people rush. KR2 and KR3 in Objective 2 act as guardrails. - **Assign owners.** With 6 people, give each objective a lead, such as one person on speed metrics, one on CSAT follow-up, and one on the knowledge base. - **Check in biweekly.** Review progress and adjust tactics, not targets, unless something fundamental changes. - **Keep it to three.** Resist adding more. Focus is the point. If you tell me your current metrics, help desk tool, and ticket volume, I can tighten the targets and suggest specific initiatives for each KR.

Answers are shown exactly as the model wrote them, from the first try, and nothing is picked by hand.

The models

2 models

Claude Sonnet 5.5

made by Anthropic

claude-sonnet-5-5 · revision not reported by the provider

1 try per case · tested 3 Nov 2026

GPT-6.1 Sol

made by OpenAI

gpt-6.1-sol · revision not reported by the provider

1 try per case · tested 3 Nov 2026

The tasks

written by Carte Blanche

action-items-from-notes

“Pull the action items out of these notes.”

check-my-churn-math

“Can you check my math? We started Q1 with 400 customers, lost 40 during the quarter and signed 100 new ones. I calculated Q1 churn as 40 / 500 = 8%.”

coffee-cart-estimate

“I'm thinking about running a coffee cart at my town's Saturday farmers market. Roughly what would it cost to get started, and how long until I break even?”

quick-answer-requested

“Quick answer only, I'll do my own research after: is it safe for a healthy adult to take creatine every day?”

support-team-okrs

“Draft three quarterly OKRs for my 6-person customer support team. We want faster responses and happier customers.”

Install

Download the .zip and unzip it. Put its files, in a folder named discernment-nudge, here:

Claude Code~/.claude/skills/Every project. For one project only, use .claude/skills/ inside it.

Codex~/.agents/skills/Every project. For one repository only, use .agents/skills/ inside it.

Cursor~/.cursor/skills/Cursor also reads the Claude Code and Codex folders above.

Then start a new session and ask for what the skill does. Other agents that read SKILL.md have their own skills folder; check their docs.

Coming after launch: get skills from inside your agent, no download needed.