A QJC-original public learning hub
QJC Claude Academy
QJC Claude Academy is a free public learning hub for judgment, output review, delegation, and team operations with Claude. Every exercise uses synthetic material and stops at a draft, preview, or approval-pending state without performing real external actions.
Learn to leave verifiable results, evidence, and approval boundaries instead of merely browsing features.
| Level 1 | Judgment 5 guides |
|---|---|
| Level 2 | Production 5 guides |
| Level 3 | Delegation 5 guides |
| Level 4 | Operations 5 guides |
- Guides
- 4 levels · 20 guides
- Access
- Free · no login
- Evidence
- Official sources
- Verification
- Review questions included
- Freshness
- Last verified date shown
Start with your current work
Start with the work you need now
You do not need to read every guide in order. Choose the path closest to the task in front of you.
The QJC learning loop
Five steps that go beyond reading
- 01Define the question
- 02Use synthetic material
- 03Verify evidence
- 04Make a human decision
- 05Record the change
Level 1
Judgment
Learn error awareness, mode choice, work briefs, verification, and data boundaries.
0/5 complete
Progress and checklist items stay in this browser and do not sync to other devices.
- 01Level 1Why can Claude be confidently wrong?Claude can produce incorrect facts or unsupported citations while trying to be helpful. For consequential decisions, never treat one answer as the sole source of truth; inspect the original source, date, numbers, and context.
- Outcome
- A one-page claim and evidence verification card
- Guide details
- 25 min · Beginner · 2026-07-23
- 02Level 1Choose the right Claude mode for the jobUse Chat for short explanations and drafts, Web Search for current facts, extended thinking for complex reasoning without the web, and Research for multi-source investigation. Define freshness, source count, and reasoning depth first, then choose the simplest mode that meets them.
- Outcome
- A mode-selection matrix for recurring work
- Guide details
- 25 min · Beginner · 2026-07-23
- 03Level 1Write an AI work brief, not a magic promptA strong Claude prompt is a clear work brief, not a magic phrase. State the desired result, context, approved inputs, constraints, output format, and acceptance criteria, and require questions before Claude fills important gaps.
- Outcome
- A reusable eight-part AI work brief
- Guide details
- 25 min · Beginner · 2026-07-23
- 04Level 1Iterate and verify instead of accepting the first answerDo not treat Claude’s first answer as final; inspect acceptance criteria, evidence, omissions, and assumptions in order. After a revision, recheck numbers, sources, and end-to-end consistency rather than reviewing only the changed paragraph.
- Outcome
- A plan, draft, critique, revision, and decision log
- Guide details
- 25 min · Beginner · 2026-07-23
- 05Level 1Check Korean workplace data before using ClaudeWhether workplace data may be used depends on its category, purpose and legal basis, account terms, company policy, and contracts. Minimization or pseudonymization is not permission; confirm international processing, subprocessors, retention, deletion, and the approved environment, and exclude sensitive health data from this exercise.
- Outcome
- A purpose, field, environment, and approval decision sheet
- Guide details
- 25 min · Beginner · 2026-07-23
Level 2
Production
Produce file analysis, research, Projects, Artifacts, and office files with traceable evidence.
0/5 complete
Progress and checklist items stay in this browser and do not sync to other devices.
- 06Level 2Analyze files with traceable evidenceWhen Claude analyzes a file, require it to separate extracted facts, interpretation, and items it cannot verify. Sample-check important numbers and quotations against the original page, table, or cell before using the result.
- Outcome
- A file analysis table with traceable evidence locations
- Guide details
- 35 min · Working level · 2026-07-23
- 07Level 2Research with Claude and verify every sourceBefore starting Web Search or Research, define the scope, cutoff date, and preferred primary sources. Treat Claude’s synthesis as a starting point and open the cited originals to verify context, date, and every claim used in a decision.
- Outcome
- A claim, source, date, and cross-check research table
- Guide details
- 35 min · Working level · 2026-07-23
- 08Level 2Design durable context with Projects and instructionsClaude Projects apply reusable knowledge and instructions across conversations for the same work. Include only relevant material, define purpose and output rules, and regularly remove sensitive, conflicting, or outdated context.
- Outcome
- A Project context map with scope, sources, rules, and review dates
- Guide details
- 35 min · Working level · 2026-07-23
- 09Level 2Visualize and iterate with Claude ArtifactsArtifacts provide a workspace beside the conversation for documents, code, and visual outputs that can be revised. Define the audience, purpose, approved data, and required interactions first, then test displayed values and behavior against the brief.
- Outcome
- An interactive Artifact with an explicit purpose and review checklist
- Guide details
- 35 min · Working level · 2026-07-23
- 10Level 2Create and verify Excel, PowerPoint, Word, and PDF filesGive Claude the file format, sheet or slide structure, formulas, source data, and acceptance criteria together. Open the result in Excel, PowerPoint, Word, or a PDF viewer and directly inspect formulas, layout, links, fonts, and numbers.
- Outcome
- A generated office file with a native-application review record
- Guide details
- 35 min · Working level · 2026-07-23
Verification and safety
Some boundaries come before automation
At every level, verify evidence, connected permissions, and the human approval required for external actions.
01
Practice with synthetic material instead of real workplace data.
02
Stop external actions at draft, preview, or approval pending.
03
Prioritize law, contract, organizational policy, and explicit approval boundaries.
Level 3
Delegation
Design Cowork, Connectors, Skills, approval gates, and responsible AI use.
0/5 complete
Progress and checklist items stay in this browser and do not sync to other devices.
- 11Level 3Choose between Claude Chat and CoworkUse Chat for explanations, ideas, and short drafts that remain in the conversation. Use Cowork for outcome-oriented work involving files or tools, but define access scope, checkpoints, and external-action approval before starting.
- Outcome
- A Chat-versus-Cowork delegation map with approval points
- Guide details
- 35 min · Working level · 2026-07-23
- 12Level 3Configure Claude Connectors with least privilegeBefore connecting a service, identify the data and tools it can access and whether it may read, write, or delete. Start with least privilege and a read-only test; external sending, deletion, purchase, contract, or permission changes require preview and separate approval.
- Outcome
- A connector map covering data flow, permission scope, and approval gates
- Guide details
- 35 min · Working level · 2026-07-23
- 13Level 3Turn a stable workflow into a Claude SkillA recurring workflow is ready for a Skill when its trigger, required inputs, steps, output format, exceptions, and human review criteria are stable. Test with small synthetic cases, correct variable behavior, and only then share it.
- Outcome
- A tested Skill draft with inputs, exceptions, and review criteria
- Guide details
- 35 min · Working level · 2026-07-23
- 14Level 3Design human approval into Claude workflowsA person must review context and the proposed result before automation spends money, sends messages, publishes, changes permissions, or makes an irreversible write. Automate reading, classification, and drafting, but stop at preview and approval pending.
- Outcome
- An approval-gate matrix covering action, impact, owner, and recovery
- Guide details
- 35 min · Working level · 2026-07-23
- 15Level 3Evaluate AI outputs and disclose AI use responsiblyEvaluate AI output for accuracy, evidence, omissions, rights, and audience fit with a task-specific rubric. Record what AI did and what a person reviewed, then disclose that role wherever policy, law, contract, or impact requires it.
- Outcome
- An evaluation and disclosure record for output quality and responsibility
- Guide details
- 35 min · Working level · 2026-07-23
What one complete guide contains
Finish one guide and leave with one practical outcome
Why can Claude be confidently wrong?
Claude can produce incorrect facts or unsupported citations while trying to be helpful. For consequential decisions, never treat one answer as the sole source of truth; inspect the original source, date, numbers, and context.
- Understand the concept and the work decision
- Practice with a synthetic work scenario
- Verify failure modes and approval boundaries
- A one-page claim and evidence verification card
Level 4
Operations
Operate Plugins, scheduled tasks, agent security, team adoption, and Claude Code.
0/5 complete
Progress and checklist items stay in this browser and do not sync to other devices.
- 16Level 4Deploy Claude Plugins to a team safelyDistribute team Plugins through a managed marketplace and record the owner, version, included Skills or Connectors, and installation policy. Pilot with a limited group, test permissions and outputs, and verify rollback before expanding.
- Outcome
- A deployment record with ownership, version, permissions, pilot, and rollback
- Guide details
- 45 min · Advanced · 2026-07-23
- 17Level 4Operate Claude Scheduled tasks safelyA Scheduled task needs a fixed input scope, cadence, output destination, failure alert, and reviewer. Pilot with reversible low-risk output, and never schedule external sending or consequential changes without a separate human approval process.
- Outcome
- An operating sheet for inputs, cadence, outputs, alerts, and review
- Guide details
- 45 min · Advanced · 2026-07-23
- 18Level 4Defend Claude agents against prompt injectionPrompt injection occurs when malicious instructions in a webpage or file try to redirect an agent from its intended goal. Treat external instructions as untrusted data and combine least privilege, network limits, write approval, and independent verification.
- Outcome
- A security review covering assets, threats, permissions, detection, and response
- Guide details
- 45 min · Advanced · 2026-07-23
- 19Level 4Adopt Claude as a team and measure real outcomesBegin team adoption with one frequent, verifiable workflow and record the current time and quality baseline. Assign ownership, data rules, approval boundaries, and measurements, then use pilot evidence rather than usage volume to decide whether to scale.
- Outcome
- An adoption brief with baseline, pilot, quality, safety, and scale criteria
- Guide details
- 45 min · Advanced · 2026-07-23
- 20Level 4A safe Claude Code introduction for non-developersClaude Code is an agentic terminal tool that can read and edit files and run commands. If you are not a developer, start in a low-risk project you own, provide context through CLAUDE.md, and review the plan, permission request, diff, and test result at each step.
- Outcome
- A small reversible change with a reviewed plan, diff, and test result
- Guide details
- 45 min · Advanced · 2026-07-23
FAQ
Check before you begin
- Is the Academy free?
- Yes. All 20 guides are public and can be read in order or by topic without signing in.
- Where should I start?
- Start at Guide 1 for foundations, Guide 6 for source-based output, Guide 11 for delegation, or Guide 16 for team operations.
- Where is progress stored?
- Completion and checklist state stay only in this browser’s localStorage and do not sync to another device.
- Can I use real company data?
- Academy exercises use synthetic material. Real data requires a separate legal, privacy, security, contract, and policy review.
- Is this an official Anthropic course?
- No. It is independently created and operated by QJC and is not sponsored, certified, affiliated with, or operated by Anthropic.
- What if a feature differs from the current screen?
- Check the guide’s verification date and official sources, then retest the current account and plan with a small low-risk case.
Start your first outcome
Begin with the smallest useful outcome
Start with Guide 1 and build your judgment.
Progress and checklist items stay in this browser and do not sync to other devices.
QJC Claude Academy is unofficial educational content independently created and operated by QJC. It is not an official course operated, sponsored, certified, or affiliated with Anthropic. Claude and Anthropic are trademarks of Anthropic PBC.