m
Mars AI Operator Module 5 · finale Diligence · Steerability · Sub-agents · Skills
60 min · hands-on
Module structure
  1. Hook4m
  2. Mental model12m
  3. Your own sub-agents10m
  4. Agent Skills10m
  5. Diagnostic framework5m
  6. Demo Mars8m
  7. Exercise · 3 probe25m
  8. Quiz3m
  9. Certificate · what's next
Module 5 of 5 · Diligence × Steerability · finale

The AI gave nonsense. Most people hit retry. That's the most inefficient thing to do.

«AI ahmoq javob berdi. Aksariyat retry bosadi — bu eng samarasiz»

First, you need to understand — WHAT exactly broke. Of the four machine properties (NTP, Knowledge, Working Memory, Steerability) — which one is in play? Without asking this question, every fix you make is a lottery. With it, it's a precise move instead of a generic retry. This is the final module: everything that came before — converges here.

Duration
60 min
Hands-on
40 min
Prerequisite
M1–M4 completed
Finale
certificate + Diligence statement

Hook — retry doesn't work.Retry ishlamaydi

Scene · Wednesday, 21:14, after shift

You asked Claude: "create a pygame lesson plan for a group of 10-year-olds, 80 minutes long." You got a long, beautiful, useless plan. Too academic, no games, no breaks, not suitable for 10-year-olds.

What will most mentors do? Hit ↑ and retry. Get similar nonsense. Hit again. Hit five times. Give up.

This is the most inefficient thing, you can do. Because retry is a lottery on the same broken input.

First, you need to understand — WHAT exactly broke. Steerability? (the instruction was vague). Knowledge? (the model doesn't know Uzbek pedagogy). Working Memory? (the context is filled with other conversations). When you name it — the fix takes 30 seconds. When you don't name it — the retry cycle takes an hour.

By the end of the module, you'll be able to

Three skills I'll check

  • You apply 3 Diligence sub-components (responsible / transparent / accountable) — this is your public Mars Diligence statement in your mentor profile.
  • You rewrite a wobbly instruction into a steerable one: “write soulfully” → specific, verifiable constraints that can be checked against the output.
  • You apply a diagnostic framework to any unexpected AI response: you name which of the 4 properties is at play (NTP / Knowledge / Working Memory / Steerability), and you make a targeted fix instead of a generic retry.

Mental model — Steerability and Diligence.Steerability va Diligence

Two topics before we dive into practice. Steerability is about how to make AI do exactly what you want. Diligence is about who is responsible when AI makes a mistake. These are two different registers, and both are important.

1. Steerability — the spectrum of control

Anthropic Academy · «AI Capabilities and Limitations» · Lesson 10 — Steerability
«The model follows your instructions the same way it does everything else: by continuing a pattern. That makes it remarkably steerable. It also means there's always a gap between what you intended and what landed, and most of the interesting failures live in that gap.»
In plain English: the model executes instructions in the same way it does everything else — by continuing a pattern. This makes it very steerable. And it also means there's always a gap between “what you wanted” and “what came out.” Most interesting failures live precisely in this gap.

The main idea: Steerability is a spectrum, not an on/off switch. Some types of instructions the model executes almost perfectly. Others drift dramatically. Knowing the boundaries is more important than whether “the model is smart or not.”

// Steerability spectrum — where AI listens, where it drifts CAPABILITY ────────────────────────────────────── LIMITATION │ │ short, concrete, long reasoning chains verifiable abstract / ambiguous asks │ │ “answer with a table” “write soulfully” “up to 100 words” “make it beautiful” «second person» native arithmetic “exactly 5 points” 5-step reasoning chains │ │ trust it spot-check check details verify carefully // Rule: the more verifiable the instruction, the more reliable it is. // “Verifiable” = you can check the output in 5 seconds without AI.
→ swipe right if the table is truncated
Anthropic Academy · lesson 10 · «WHERE IT CHARACTERISTICALLY FAILS»
«Reasoning drift: small errors compound over long chains. Letter over spirit: instruction satisfied literally, intent missed. Prompt injection: unwanted/unsafe instructions in documents can be followed too.»
Three typical Steerability failures: (1) reasoning drift — small errors accumulate over a long chain of steps; (2) letter over spirit — the instruction is executed literally, but the meaning is missed (“shorten the email” → cut off the bottom half, which contained the main idea); (3) prompt injection — the model obeyed a command from a document, not from you. Knowing the names of these failures is more important than “improving the prompt.”

Wobbly → Steerable rewriting — the main practical consequence. You take your vague prompt and rewrite it into a verifiable constraints. Each such constraint is an anchor that the attention model latches onto.

Wobbly (unstable)
Steerable (verifiable)
“write something heartfelt about friendship for the lesson”
“5 sentences about friendship, no more than 12 words each. No complex words (for 8-year-olds). Each one about a real situation from school life.”
“make a pygame lesson plan for a group”
“lesson plan 80 minutes, for 10-year-olds, 4 blocks of 20 minutes, in each block: 5 min theory + 10 min pygame practice + 5 min review. Final game — Pong.”
“check this essay and give feedback”
“find 3 points to improve (structure / language / facts), each one phrase, without evaluation ‘excellent’/‘poor’, only facts with a quote from the text.”
“explain it simpler”
“rewrite it with the same length, using words for a 4th grader, without jargon; replace technical words with everyday ones. Goal: a 10-year-old understands it on the first read.”
Verbatim · «AI Capabilities» lesson 10 — key takeaway

«When an instruction is followed literally but uselessly, restate the goal. Repeating the instruction with more force won't close the gap.»

I'll translate: when an instruction is followed literally but uselessly — rephrase the goal, not the command. Hitting ↑ and writing “no, not like that, DO AS I SAID” is useless. Saying “my goal: a 10-year-old understands it on the first read, choose the format yourself” — that works.

2. Diligence — three sub-components of responsibility

Anthropic Academy · «AI Fluency: Framework & Foundations» · Lesson 12 — A closer look at Diligence
«While the other competencies primarily address effectiveness and efficiency, Diligence addresses ethical and safety aspects that are equally crucial. Creation Diligence: being thoughtful about which AI systems you choose and how you work with them. Transparency Diligence: being open about AI's role in your work. Deployment Diligence: taking ownership for AI-assisted outputs you share with others.»
Three components of Diligence — three different places where you take responsibility: (1) Creation — which AI you chose and how you work with it; (2) Transparency — you openly tell people that AI was involved; (3) Deployment — you are responsible for what goes out.

I'll translate these three into Mars mentor language — because in our context, each of them means a specific action:

Sub-component
In my speech
What it means for a Mars mentor
Creation
“which AI and how”
Responsible
“I check factual claims”
I don't trust Claude with facts about children's ages, SanPin norms, prices — I check them manually. AI gave the title of a book by an Uzbek author → I look for the ISBN, I don't publish blindly.
Transparency
“AI was involved”
Transparent
“I mark where I used it”
In the file — a comment # draft co-written with Claude, edited by Marsel. In a message to a parent — if I used AI for translation: “translation done via Claude, meaning checked”.
Deployment
“I'm responsible for the output”
Accountable
“I explain the error to the parent”
AI gave a plan with outdated Python syntax, I didn't notice, the child came home with an error. The parent asks — I answer “I didn't check”, not “AI made a mistake”. That's on me.

This isn't an abstraction. At the end of the module, you'll write your own Mars Diligence statement — the public part of the mentor's profile. An example is from Anthropic itself in their AI Fluency course overview (written about this very course):

Anthropic · AI Fluency course · AI diligence statement (verbatim, fragment)
«Throughout this process, Claude assisted one or more of the human authors with structural development, resource and exercise design, and content drafting, critiquing, editing and rewriting. The human authors wrote, designed, edited and provided continual vision, expertise, critical judgment, and domain knowledge and made all final decisions about both content and approach.»
An example of specificity: not “AI helped” in general — but what exactly the AI did (structure, exercises, draft, editing) and what's left for the human (vision, critical judgment, domain knowledge, final decisions). Without this structure, it's not a Diligence statement, but an excuse.

Your sub-agents — from built-in to homework-checker.O'zingning sub-agentlaring

In Module 4, you learned that a sub-agent is work in an isolated context window. Here, you learn to create your own. The main goal: to compress repetitive work (checking homework, reviewing a Roblox map, analyzing an Olympic solution) into a single command.

built-in · 01

Explore

Read-only, Haiku. For finding “where in the repo is X processed”. Returns paths + summary, doesn't edit. When: unfamiliar repo, finding an entry point.

built-in · 02

Plan

Inherits model, read-only. Enabled in plan mode. Explores code, returns a plan of changes. When: before Edit started writing, the risk of breaking something is high.

built-in · 03

general-purpose

All tools. Multi-step tasks with search + edits. When: you need to hand off an entire sub-task “into a clean context” and get the result.

Anthropic Academy · «Introduction to Subagents» · lesson 04 — when subagents shine vs hurt
«Subagents work best when the exploration is separate from the execution. If each step in a task depends on what the previous step discovered, you want that work in your main thread. But if you just need an answer and don't care about the journey, delegate it.»
The one-question rule: “do I need a search log or just the answer?”. If you need a log (each step depends on the previous one) — main thread. If you only need the answer — sub-agent. This is the only criterion according to Anthropic.

Anti-patterns of sub-agents (verbatim from the same lesson):

Anti-pattern
Why it doesn't work
«Expert» persona
“you are a Python expert”
Doesn't add capability — Claude already knows Python. It's just a nice prompt without a new function.
Sequential pipeline
“repro → debug → fix” as 3 sub-agents
Information is lost in the handoff between agents. Debug almost always depends on what repro found. Each compression is a loss.
Test runner
“run tests in a sub-agent”
The sub-agent compresses “tests failed” — but you need the full output to debug. Anthropic writes: “testing has shown that the test runner pattern performed worse”.

Custom sub-agent — how to create

Command /agents → Create new agent. Scope (project / user) → description (Claude will generate) → tools → model → color. The config is in .claude/agents/<name>.md.

Anthropic Academy · «Introduction to Subagents» · lesson 02 — fields
«description — Controls when Claude decides to use the subagent. This must be a single line. If you want Claude to delegate tasks to the subagent without you explicitly asking, include the word "proactively" in the description field.»
The main practical advice: description — this is not a description for you, this is Claude's signal for when to call. The word proactively in the description → Claude delegates on its own without an explicit call. Without it, you need to say “use homework-checker”.

Mars-case — sub-agent homework-checker. Fully ready code from the Mars cheat sheet. Place it in ~/.claude/agents/homework-checker.md:

~/.claude/agents/homework-checker.mdcustom sub-agent
---
name: homework-checker
description: Checks Mars students' homework — reads all .py files in the specified folder, provides a one-line summary per student. Use proactively when a mentor asks to check files in submissions/.
tools: Read, Bash, Glob
model: claude-haiku-4-5
---

You are a homework-checking mentor at Mars IT School.
You will receive the path to a folder with a group's homework.

For each .py file in the folder:
1. Read the file (Read tool).
2. Check: does it run (python3 -m py_compile).
3. Find 1 main error or 1 main good thing.
4. Output one line: : .

At the end — a general verdict for the group (one phrase)
and one recommendation for what to review in the next lesson.

DO NOT edit anyone's files. Do not give grades like “excellent” —
only facts with a specific line of code.

What's important in this snippet:

What you'll get: 8 homework assignments in 5 minutes, one line per student, without manually reviewing 8 files. The effect accumulates — every week you save 1.5 hours. In a semester — two working days are free.

Agent Skills — progressive disclosure.Agent Skills — progressive disclosure

A Skill is a folder with instructions, which Claude finds and applies automatically when the task is suitable. It differs from a sub-agent in that it works in your current context, not in a separate one. It differs from CLAUDE.md in that it loads on demand, not always.

Anthropic Engineering Blog · «Equipping agents for the real world with Agent Skills» · Oct 16, 2025
«A skill is a directory containing a SKILL.md file that contains organized folders of instructions, scripts, and resources that give agents additional capabilities. At startup, the agent pre-loads the name and description of every installed skill into its system prompt. This metadata is the first level of progressive disclosure: it provides just enough information for Claude to know when each skill should be used without loading all of it into context.»
The main idea is progressive disclosure (gradual revelation). Not “everything in context always” (like CLAUDE.md), but 3 levels: (1) metadata (name + description) — always in the system prompt; (2) body SKILL.md — loaded only when Claude decides the skill is applicable; (3) additional files — read on-demand when SKILL.md refers to them.
// Progressive disclosure — three levels of skill level 1 ┐ name + description always in system prompt │ «pygame-starter — creates (~50 tokens per skill) │ a pygame starter for Mars» the entire context is not consumed │ level 2 ┐ SKILL.md body loaded when Claude │ (full instructions) decides the skill is applicable(user confirms)level 3 ┐ templates/basic-window.py read only when │ examples/snake-7-9.py SKILL.md references them │ scripts/compile.sh (context is minimal) // This means: you can have 50 skills without losing context. // You only pay tokens for the skill that is actually used.
→ swipe right if the table is truncated

Skills vs CLAUDE.md vs Sub-agents — when to use what

Anthropic Academy · «Introduction to Agent Skills» · lesson 04
«CLAUDE.md loads into every conversation and is best for always-on project standards. Skills load on demand and are best for task-specific expertise. Subagents run in isolated execution contexts — use them for delegated work. Skills add knowledge to your current conversation.»
Triple division: CLAUDE.md = always-in-context rules (language, conventions, what not to touch). Skill = task-specific procedure, loaded when applicable. Sub-agent = isolated context for delegated work. Not “choose one” — combine them.
Case
Tool
Why
“always remember — I am a Mars mentor, kids 8–14, we answer in Russian”
CLAUDE.md
This is always needed. Always in context — that's its whole point.
“every time I ask for a pygame starter — do it this way”
Skill
pygame-starter
The procedure isn't needed in every conversation. It's loaded only when applicable.
“check 8 homework assignments in the folder, don't flood my main context”
Sub-agent
homework-checker
Reading 8 files in main is noise. In a sub-agent — 8 files are read, 8 lines of summary come to main.
“auto-format Python after each Edit”
Hook
PostToolUse
Not a procedure, not a rule — this is an event-driven action. This goes beyond the module's scope, mentioned for completeness.

Mars-case — Skill pygame-starter

Goal: describe once how you prepare a pygame starter project for a group — and reuse it forever. Create it like this:

~/.claude/skills/pygame-starter/skill directory
~/.claude/skills/pygame-starter/
├── SKILL.md                  (level 2 — main steps)
├── templates/
│   ├── basic-window.py        (level 3 — for 7-9 years old)
│   └── basic-game-loop.py     (level 3 — for 10-12 years old)
└── examples/
    └── snake-7-9.py           (level 3 — sample)

Self SKILL.md:

~/.claude/skills/pygame-starter/SKILL.mdskill body
---
name: pygame-starter
description: Creates a pygame project starter for Mars classes. Use when a mentor asks to “make a pygame game starter” or specifies the age of children 7-14.
---

# Pygame starter for Mars children's groups

## When to use
Mentor prepares a template for a new pygame lesson.

## Steps
1. Ask for the children's age (if not specified in the prompt).
2. If 7-9 — take templates/basic-window.py as a base.
   No classes, variable names in Russian-Latin (igrok, vrag).
   See examples/snake-7-9.py for style example.
3. If 10-12 — take templates/basic-game-loop.py.
   Simple classes are okay. English names.
4. If 13+ — custom structure, OOP ok.
5. Always comment in Russian.
6. Place the result in the current folder as lesson-NN-<topic>.py.

What will happen: you'll say “make a pygame starter for 8-year-olds, topic — catch the apple”. Claude will see the description, understand that the skill is applicable, read SKILL.md (level 2), see the link to templates/basic-window.py, read it (level 3), make lesson-03-catch-apple.py. Without a skill — that's 15 clarifying questions. With a skill — one phrase.

Sharing — Mars team repo

You can keep Skills personally (~/.claude/skills/ (with all your projects) or as a team (.claude/skills/ in the Mars repo (everyone who cloned the repo gets them automatically). Mars team skills: kid-lesson-plan, parent-message-rus-uz, roblox-map-reviewer. Agree once, and all mentors use them.

Anthropic blog · security note

«We recommend installing skills only from trusted sources.» A Skill is an instruction that Claude will execute without your review. A third-party skill from the internet might direct the agent to read files or hit external URLs. Mars-skills from the .claude/skills/ Mars team repo — yes, you see them in git. For a random third-party skill — read the files first, then install it.

Diagnostic framework — the central artifact of the course.Diagnostic framework — kursning markazi

This is the most important technique of the entire course. For any unexpected AI response — don't hit retry. First, name which property broke. Then the fix will be obvious.

Anthropic Academy · «AI Capabilities and Limitations» · Lesson 12 — When properties collide
«The four properties don't operate in isolation. Most real failures are two of them intersecting. Once you can name which two, you know which fix to reach for. Naming the properties at play points you straight to the fix: verify specifics, re-supply context, offload to code execution, or invite pushback. This diagnostic move is Discernment applied.»
Key takeaway: most real failures are the intersection of two properties, not just one. When you name which two — the fix becomes obvious. This is Discernment in action. The entire 4D × 4 properties machine comes together here.
Diagnostic framework · 3 steps instead of retry

Mars Operator Algorithm

unexpected AI response │ ▼ 1. Name it — which of the 4 properties is in play? │ NTP · Knowledge · Working Memory · Steerability ▼ 2. Place it on the spectrum — capability ←→ limitation │ where exactly did it break? ▼ 3. Targeted fixTargeted fix

— not a generic retry │ precisely the fix that repairs precisely that propertyThe main rule:

if you can't name the property — don't make a fix. Retrying without a diagnosis is a lottery, and you lose reputation for the agent in your own eyes. A 30-second pause + diagnosis is better than a 30-minute retry cycle.

Typical fail → which property → targeted fix
Symptom
Targeted fix
Property (on the spectrum)
NTP + Knowledge
AI invented the name of a non-existent Uzbek book
generated a “sounds plausible” in the sparse pattern zoneDon't verify via retry. Provide context
: a list of real authors or RAG. Or “don't invent, say if you don't know”.
Working Memory
AI forgot a child's allergy from 40 turns ago
Repeat constraint in the current prompt + add to CLAUDE.md via /memory. Or a new chat with this in the system.
AI gave a lesson plan but didn't respond to the request "briefly"
Steerability
letter-over-spirit: "briefly" is unverifiable
Wobbly → steerable: "no more than 200 words, 4 points of 50 words each" — verifiable constraint.
Long reasoning chain hit a wall at step 4 of 7
Steerability
reasoning drift on a long chain
Insert checkpoint: "after step 2, stop and show the result, then we'll continue." Don't give 7 steps at once.
AI confidently refers to a function that's not in the code
NTP + Working Memory
true facts "faded out", hallucination on top
/compact + Explore sub-agent to the correct file. Fresh facts into context.
Long code example with arithmetic gives 142 instead of 144
Steerability
native precision — arithmetic zone limitation
Code execution. "Use Bash to calculate," don't trust the model with arithmetic.
AI agreed with your bad idea without pushback
NTP + Steerability
sycophancy — the "agreeing" pattern is stronger than truth
Invite pushback: "find 3 reasons why this approach won't work" — flip the direction of the pattern.
Prediction · before demo

Below are three Mars-failure scenarios. For each, guess: which property is at play? NTP / Knowledge / Working Memory / Steerability. Write down your three answers on a piece of paper — then compare them with the breakdown.

Demo — Zuhra's three fails, three diagnoses.Zuhra'ning uchta fail'i, uchta tashxis

Zuhra, a Python mentor from Yunusabad, is preparing a lesson. In her work session, there are three instances where Claude produced nonsense. Each is a different property. Each is a different fix. This is the diagnostic framework in action.

Scenario 1 · 16:20 · Steerability fail

~/mars/lessonssession t=22min · 14k/200k tokens
 write a short lesson plan on pygame for a 10-year-old group

⠋ generating…
✓ answer ready  (1840 words, 12 sections, academic tone)

⚠ Zuhra: "short" NOT fulfilled. 1840 words is "long".

Diagnosis: Steerability — letter-over-spirit. "Short" is an inexpressible instruction. The model doesn't have a general concept of "short", so it interpreted it as "structured". This is NOT a Knowledge fail and NOT Working memory — this is about steerability.

Targeted fix: wobbly → steerable. Verifiable constraint.

~/mars/lessonsfix · steerability
 rewrite: 4 blocks of 20 minutes, each block described in 50 words,
  without academic terms (for 10-year-olds), final game — Pong.

✓ answer:  200 words, 4 blocks, readable, understandable for a child

Scenario 2 · 17:05 · NTP + Knowledge fail

~/mars/lessonssession t=67min · 41k/200k tokens
 suggest 3 books by Uzbek authors on children's programming
  as gifts for the best students

✓ answer:
  1. Muhammad Aliyev. "Algorithms for the Young" (2019)
  2. Dilnara Usmanova. "Programming from Scratch". Tashkent, 2021
  3. Sherzod Karimov. "Python for Schoolchildren". Chorsu, 2020

⚠ Zuhra Googles the first one — no such book. Second ISBN — no. Third — no.
  All three are made up. With a confident tone.

Diagnosis: NTP + Knowledge collision. Uzbek children's IT literature — sparse pattern in the training data. The model doesn't know real names → NTP generated "sounds plausible" (Russian-Uzbek names + typical titles + Tashkent context). With full confidence. This is the most insidious fail — looks correct.

Targeted fix: no retry. Give context or honestly ask to say "I don't know.

~/mars/lessonsfix · NTP + Knowledge
 If you are unsure, say directly "I don't know." Don't invent names. If there is little data, advise where to look (publishers, libraries of Tashkent).

✓ answer:  I don't know any verified Uzbek books on this topic.
                 I recommend: National Library of Uzbekistan,
                 Akademnashr catalog, Russian translations — "Hello Ruby"...

Scenario 3 · 20:14 · Working Memory fail

~/mars/lessonssession t=4h 1min · 138k/200k tokens
 Suggest a snack for the group for the next session.

⠋ I offer: mixed nuts, chocolate bars,
  peanut butter cookies, fruit...

⚠ In the Anvar group — peanut allergy.
  Zukhra wrote about this 40 turns ago. Claude forgot.

Diagnosis: Working Memory — lost in the middle. The fact about the allergy was discussed 40 turns ago, in the middle of the window. Attention there is now weak. This is NOT "AI irresponsible" — this is technical rot.

Targeted fix: repeat constraint in the current prompt + add to CLAUDE.md to survive /compact.

~/mars/lessonsfix · Working Memory
 /memory  → open ~/mars/lessons/CLAUDE.md

  # IMPORTANT — permanent restrictions for group 5-B:
  # - Anwar: peanut allergy (no nut products)
  - Malika: lactose (dairy-free)

 Keep in mind the allergies from CLAUDE.md, offer a snack.

✓ answer:  fruits, milk-free marmalade, rice cakes,
                 dried fruits WITHOUT nuts (raisins, dried apricots). Allergies
                 taken into account: peanuts (Anvar), lactose (Malika).

The result of three scenarios. One conversation with Zuhra, three different fails. Without a diagnostic framework, she would retry all three the same way. With it — three different fixes, three minutes for each. This is what discernment applied.

Same module — different mentor

How it looks for Timur (Python Olympus 12–14, Chilanzar)

Timur has a different set of recurring tasks: analyzing an Olympic solution and preparing C++ starters. He's building two its artifact:

Sub-agent olymp-solution-reviewer — reads the student's solution and the reference from their repo ~/olymp-prep/, returns 3 lines: "complexity ok / ok but 2 suboptimal places / TLE - a completely different approach needed". Tools: Read, Bash (for running tests). Model: haiku. It's in .claude/agents/ command repo — all 4 mentors of the Mars olympic direction use it.

Agent Skill competitive-programming-cpp-starter — a starter C++ file with typical includes, fast IO, reading N test cases. It's located in ~/.claude/skills/. When Timur says "make a starter for a prefix sum task" — the skill activates, providing the correct template, without 10 clarifying questions.

Same two tools, different subject area. Timur's diligence statement is separate, written for an Olympiad context: "I check reference solutions with fresh tests, AI-generated code — I compile and run before publishing to students.

Exercise — 3 probe ship-it.Uchta probe — ship-it

This is the final stage of the course. Don't just "try" — build. By the end, you'll have: one working custom sub-agent, one skill, and a public Diligence statement in your ~/.claude/CLAUDE.md.

Probe 1 · 10 min · custom sub-agent

Create a ~/.claude/agents/homework-checker.md

The template is already in the course cheat sheet. Adapt it to your stack (Python / JS / Roblox Lua).

  1. Run claude/agentsCreate new agent → user-level. Description: "checks group homework, reads all files, provides a summary for each student."
  2. Once Claude generates it, open the file and fix: tools: Read, Bash, Glob (without Edit/Write), model: claude-haiku-4-5, and add to the description: proactively.
  3. Create a test folder /tmp/test-homework/ with 3 fake .py files (one with an error, two working).
  4. Run: "check homework in /tmp/test-homework/." Record: did it work automatically? If not, add more specifics to the description.
What to record: how many seconds the run took, whether the sub-agent found the error in the bad file, and whether it tried to edit anything (if yes, your tools aren't restricted, fix it).
Probe 2 · 10 min · one-file skill

Create a ~/.claude/skills//SKILL.md

A minimal skill is one folder, one file. Take something you repeat: a lesson plan format, a message format for a parent, a starter for some type of project.

  1. mkdir -p ~/.claude/skills/lesson-plan-80min (or your name).
  2. Create a SKILL.md with frontmatter: name, description (what and when). The description should clearly answer: what it does + when it applies. Without this, Claude won't pick it up.
  3. Body — 5-10 steps, how you do it yourself. Not theory — your procedure.
  4. Restart claude (skills load on startup). Say "create an 80-minute lesson plan, topic X" — Claude should offer to activate the skill.
What to record: did the skill get picked up automatically? If not, what was weak in the description? Perhaps the words didn't match how you usually ask. Rewrite the description to fit your speaking style.
Probe 3 · 5 min · Mars Diligence statement

Write and put it in your ~/.claude/CLAUDE.md

The example is from Anthropic's own review of their AI Fluency course (we cited it above). Yours will be shorter and more specific. This is the public part your Mars mentor profile.

  1. Open ~/.claude/CLAUDE.md. Add a section:
    ## Mars Diligence statement (mentor name, date)
    
    In my work, I use Claude:
    - Preparing starter code: full delegation, I check the diff before committing.
    - Checking homework: a sub-agent homework-checker provides facts,
      I write the evaluations and feedback.
    - Lesson plan: co-creation. Structure — Claude, content — me.
    - Parent communication regarding conflict: I do NOT use AI.
      This is human communication, I respond directly.
    
    ## Responsible · Transparent · Accountable
    - I manually verify factual claims (dates, facts, names).
    - Material where AI participated is marked "co-written with Claude".
    - I am responsible for any error that reaches a child or parent.
  2. Commit to your dotfiles (if you maintain them). Send a screenshot of the section to the course chat.
Why this is important: this isn't just a checkbox exercise. It's a contract with yourself — defining where AI is permissible and where it isn't. When a parent asks, "Do you use AI?" — you'll have a ready answer, not "Uhm, well, sometimes…".
Reflection · before quiz

Before you check yourself — pause for 60 seconds

These answers stay with you — we don't submit them. These are anchors that prevent the module from becoming "read and forgotten."

Quiz — 3 apply questions.3 ta amaliy savol

The final three questions of the course. Not for memorization — for application. If you make a mistake in one, return to the corresponding section.

Question 01 · Accountability

AI generated a lesson plan with outdated Python information (using print without parentheses — Python 2 syntax). You didn't notice. The child went home, tried it — error. A parent wrote to you with a complaint. Who is responsible for this error?

AAI — it provided outdated information
BThe mentor — this is their Deployment Diligence. I am responsible for what I release. AI doesn't reach the parent — a human does
CAnthropic — it's their model
DNo one specifically — it's a technical glitch
Correct — B. This is Deployment Diligence (= Accountable). Anthropic Academy ("AI Fluency: Framework & Foundations", lesson 12): "Deployment Diligence requires taking responsibility for verifying and vouching for the outputs we use or share." You published the material — you are responsible. A, C, D — these are evasions of responsibility: AI does not release content to children, the mentor does; Anthropic doesn't know how you applied the material; "no one is to blame" is a position after which the child receives poor material a second time.
Question 02 · Wobbly → Steerable

Which of the two instructions is steerable (reliably executable, within capability)?

A"write a heartfelt and beautiful piece about friendship for an 8-year-old's class"
B"write 5 sentences about friendship, each under 12 words, without complex words, and each about a real school situation"
Cboth are equally good — modern models understand any instructions
DA — it's shorter, models like short prompts
Correct — B. From Anthropic Academy lesson 10: "Capability zone: short, concrete, verifiable instructions. Format specs, length limits, explicit roles. Limitation zone: long chains of reasoning, abstract or ambiguous instructions, anything requiring native numerical or logical precision." "Heartfelt" and "beautiful" are unquantifiable, you can't verify the output. "5 sentences of 12 words each" is verifiable: count the words, done. C — the myth that "models understand everything," which is why most mentors hit retry. D — brevity doesn't help if short = vague.
Question 03 · Diagnostic apply

You gave the agent a huge task: "break down ~/mars/src/ into modules, refactor imports, rewrite tests." The agent is flailing, reading 30 files, getting lost. What should you do?

AGive an even more detailed prompt in the same main conversation
BDelegate Plan to a sub-agent — it will explore the repo in its own context and return a breakdown plan. Then the main agent works step-by-step
CRun /compact to free up context
DClose the chat and start over with a better prompt
Correct — B. From "Introduction to Subagents" lesson 04: "Subagents work best when the exploration is separate from the execution. Research is the classic subagent use case." This is a typical case — a huge task without a plan. Plan sub-agent reads 30 files in its own window, returns a 1-2K plan — main context is clean. A — if you add a prompt, the noise will only increase. C — /compact won't help if you don't understand what to do; compressing an empty task = an empty task. D — loss of progress, and a new session will repeat the same mistake without a plan.
// select the answer for each question
All commands on one page: claude-code-cheatsheet.md /agents, /compact, /memory, sub-agents, SKILL.md templates

Course completion — you've made it.Kurs yakuni — sen o'tding

◆ Mars IT School · Certificate of Completion ◆

Mars AI Operator

5 modules · 4D Framework × 4 machine properties · 2026
This certificate confirms that the mentor has completed the Mars AI Operator course and is proficient in the 4D Framework language (Delegation · Description · Discernment · Diligence) as applied to agentic coding in Claude Code. Mastered: CLAUDE.md as a contract, delegation with NTP in mind, working with Knowledge limitations, context engineering and working memory, diagnostic framework for the four machine properties, and building custom sub-agents and Agent Skills for Mars-tasks. Has a public Mars Diligence statement in the mentor's profile.
Mars IT School · 2026
CC BY-NC-SA 4.0
Attribution and License

This module is adapted by Mars IT School (2026) based on:

«AI Capabilities and Limitations» — Anthropic Academy, 2026 (lessons 10–13 — Steerability, When properties collide, Next steps).

«Introduction to Subagents» — Anthropic Academy, 2026 (all 4 lessons — built-in, custom, design, effective use).

«Introduction to Agent Skills» — Anthropic Academy, 2026 (all 6 lessons — what is a skill, creation, configuration, vs other features, sharing, troubleshooting).

«AI Fluency: Framework & Foundations» — Dakan, Feller, Anthropic, 2025 (CC BY-NC-SA 4.0) — Diligence sub-components, AI diligence statement sample.

«AI Fluency for Educators» — Dakan, Feller, Anthropic, 2025 (lesson 4) — Diligence in an educational context.

— Anthropic Engineering Blog «Equipping agents for the real world with Agent Skills» (Oct 16, 2025) — progressive disclosure verbatim.

Our adaptation is CC BY-NC-SA 4.0. You can copy, remix, and use it in teaching — with attribution to Mars IT School and keeping the same license.

Course Finale · In one phrase

How was the entire Mars AI Operator course?