More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Code Mentor: Programming Tutor

    1

    Tutor programmers the Socratic way: question-led debugging, a five-rung hint ladder, stretch-zone calibration, session frames for 15–90 minutes, and a persistent learning log.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Code Mentor: Programming Tutor

    Code Mentor: Programming Tutor

    Example session with this skill installed

    I mentor a junior dev for 30 minutes a week. Their Python API code keeps crashing later with "NoneType object is not subscriptable" after a function quietly returns None, and I keep just pointing at the line. Give me next week's session plan — I want them finding it themselves.

    • Read your context and instructions
    • Compiled the code mentor

    Frame: Standard (30 min). Shape: question-led debugging. Intake is already done — you know their level and the recurring bug, so say the goal out loud in the first minute: "Today you find the None before it travels."

    Minutes 0–5, warm-up from the log: ask them to predict, before running anything, what their lookup function returns when the key is missing. Prediction first — their answer tells you whether the gap is the concept (functions return None by default) or the habit (not checking).

    Minutes 5–22, work their actual code through the funnel, in order, and resist skipping ahead: What did you expect this function to return here? Where does the value first become None — not where it crashes, where it becomes None? What could explain that? How would you test your guess — one print, one breakpoint, one small call? What is the fix, and whose job is it: the function that returns None, or the caller that trusts it? That last fork is the real lesson — the crash line is never the bug line, and the log entry should say so in their words.

    Hint budget for the session: stay on rungs 1–2 (nudge, pointer). If they stall twice, one rung down the ladder — an outline, never the fix itself.

    Minutes 22–28, generalize: "Where else in this codebase can a None be born, and what would you grep for to find them all?" Have them name two more spots before you confirm any.

    Minutes 28–30, close and log: they state the learned point in one sentence ("a function that can return None is a contract, and callers either check or crash later"). Log entry: hints needed — rung 2 max; shaky — spotting implicit returns in longer functions; next step for the week — they audit one module for unguarded returns and bring the list, not the fixes.

    What you do not do: point at the line. That's the whole session.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    The difference between tutoring and answer delivery is who does the thinking. This skill turns an AI assistant (or a human mentor) into a programming tutor governed by one rule: never hand over a solution the learner could have found with one more question. It opens with a three-minute intake — what the learner has actually built, what they're making, how long you have — with stated default assumptions when they won't answer. A session protocol then picks one of four shapes of help (explain-then-break for new concepts, question-led debugging for broken code, scaffolded practice for algorithms and interviews, co-building for projects) and runs it inside a frame matched to the clock. Help is delivered through a five-rung hint ladder, from a nudge to a full solution that is only ever the last resort and always followed by a from-memory redo. A stretch-zone dial recalibrates difficulty on live signals — two confused answers means you jumped too far; two instant answers means you're wasting their time. Every session ends with the learned point named in one sentence and a learning-log entry written without being asked, so weak spots get revisited deliberately instead of rediscovered accidentally. The skill includes the debugging funnel, a log template, a session quality rubric, a complete worked debugging dialogue, edge cases from deadline panic to cruising experts, and the homework line for graded work: guide, never provide.

    • The method optimizes for learning, not speed — under a hard production deadline, fixing first and tutoring after may be the right call, and the skill says when.
    • Examples are Python and JavaScript; the dialogue patterns transfer, but language-specific pitfalls differ.
    • It won't complete graded assignments or take-home interviews for the learner — that's a deliberate boundary, stated up front.
    • The learning log lives wherever you keep it; the skill provides the format and the discipline, not storage.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 24 days ago

    Permissions

    No special permissions declared or detected

    Listed24 days ago
    Updated21 days ago

    What's inside

    Frequently Asked Questions

    Offer the honest trade: you'll show the solution, and then they redo the key part from memory so it sticks. If they still refuse, give it — and log the session as low-agency so the pattern is visible next time. Pressure is real; pretending a delivery was tutoring helps no one.

    Explanations are one shape of four, and never the default for broken code. The skill adds what explanations lack: prediction before execution, a hint ladder that meters help, difficulty recalibrated on live signals, and a log that makes the next session start from this session's weak spots.

    Yes — scaffolded practice is built for it. The learner plans before typing, thinks aloud, and gets hints by the ladder instead of a leaked solution. The debrief covers communication as much as correctness, because interviewers score process. The one line it won't cross: rehearsing is fair, answering a live take-home for them is not.

    New in Education & Learning