Check the input, then compare the complete deliverables
Explore the original materials, item-by-item outcomes and full report below. Attempt the input independently before using the reference to find omissions or disagreements.
Repository snippets are synthetic. No cloud agent, real code change, PR, push or deployment was performed. The exercise stops at a local task brief.
Original synthetic material. Try the input independently, then compare the reference results.
Original synthetic input
{
"synthetic": true,
"case": "cursor-cloud-agent",
"data": {
"request": "Change annual discount to 20%; FAQ explains the calculation from 12 monthly payments. Monthly price stays unchanged.",
"files": [
{
"path": "pricing.json",
"content": {
"monthly": 100,
"currency": "CNY",
"annual_discount": 0.15
}
},
{
"path": "faq.md",
"content": "The annual price is shown on the current pricing page."
}
],
"constraints": [
"Preview locally first",
"No deployment",
"No subagents",
"No push"
],
"unknown": [
"Real repository",
"Starting branch",
"Run commands"
]
}
}Processing rules
Analyze only the request and synthetic snippets. New annual=100×12×0.8=960; old annual=1020; monthly stays 100. Do not invent repository paths or commands. Cloud agents may push automatically; with no-push constraints, use a local plan and verify actual account, cost, branch and commit behavior before any later remote launch.
Complete reference results under this tutorial’s rules; nothing was written to a third-party service.Scroll horizontally within the table if needed.
| Item | Before | After | Acceptance |
|---|---|---|---|
| Monthly | CNY 100 | CNY 100 | Unchanged |
| Annual discount | 15% | 20% | Config and displayed copy agree |
| Annual price | CNY 1020 | CNY 960 | 100×12×0.8 |
| FAQ | Points to pricing page | Explains 12 months and 20% | No new renewal/refund promise |
Change the discount and matching FAQ only. Deliver actual screenshots, file changes and arithmetic checks.
No push is allowed, so this exercise does not launch a cloud agent that may push a branch.
Implementation brief — synthetic pricing request Goal: annual discount 15%→20% and matching FAQ. Acceptance: monthly 100 unchanged; annual 1020→960; configuration and copy agree. Snippets: pricing.json and faq.md; real structure unconfirmed. Missing: repository, start state, commands and pricing component. Deliver: change description, arithmetic, real desktop/mobile screenshots and check results. No real edit, agent, PR, push or deployment.
Prepare the materials and check your tools
Start with the practice files
- Download the request and synthetic configuration.
- Calculate old/new annual prices and unchanged monthly price.
- Keep repository and run-command gaps explicit.
Before using real services
- Inspect the real repository, starting state and instructions.
- Check local tools and commands; an unverified cloud setting cannot guarantee no push.
- Handle model, budget, branch, PR and push within the user’s actual scope.
Complete one file-based exercise first. For real services, verify the actual account, available tools and operation scope, then follow the acceptance check at each stage.
Download the original input and rules ↓Execution and takeover boundaries
The actions below require the agreed inputs, accounts and scope. Configuring a Routine does not expand that authorization.
Can run automatically
Analyze supplied requirements and code snippets; prepare scope, calculation comparisons and a one-shot validation plan.
Requires confirmation
Confirm repository writes, Cloud Agent launch, costs, branches, pushes, PRs and deployment with explicit scope. This exercise launches no subagent.
Take over after a failure
Do not guess files or commands without context. Stop remote work that violates constraints, preserve local diffs and logs, and let the user choose recovery.
Separate desired behavior from constraints
State what users should see, then define scope. Missing context should remain a gap rather than a fabricated component location.
Read source-data.json. List goal, allowed snippets, unchanged monthly price and unknowns. Deliver task-scope.md only. Do not inspect unrelated repositories or launch an agent.A clear change boundary.
Real repository and commands remain unknown.
Build before/after acceptance examples
Verify both the percentage and computed amount. Text-only or configuration-only changes may leave inconsistent behavior.
Calculate old/new annual prices. Compare monthly, discount, annual amount and FAQ. Keep currency and rounding assumptions; use only the synthetic plan and infer no other tiers.A price/copy acceptance table.
Old1020, new 960 and monthly 100 can be independently checked.
Write a concrete implementation brief
Include context, allowed modifications, exclusions and acceptance steps. The implementer should locate the actual files before making the smallest change.
Write implementation-brief.md: inspect real config/components, update discount and FAQ, explain 960, keep monthly unchanged, require actual screenshots and one-shot checks. Obtain commands from repository instructions; report missing commands. Add no dependencies or unrelated refactors.A reviewable implementation task.
Each requirement has a check; payment systems and other plans are out of scope.
Verify execution behavior and retain a local path
Some cloud agents push or create PRs on completion. A prompt is not a verified tool-level no-push switch.
Compare local/cloud prerequisites: account, repository, starting state, cost and push/PR behavior. Current constraints forbid push and subagents, so provide a local plan only. Do not start an agent, branch or remote task.A constraint-compliant execution plan.
No unverified cloud setting is called a guarantee.
Check actual behavior after implementation
Screenshots establish appearance, while configuration and calculations need their own checks. Use appropriate one-shot commands.
If later authorized for real local implementation, read project instructions and record the starting state. Modify only the confirmed scope. Check price, FAQ and mobile layout; report actual commands/results and real screenshots. Do not push, open a PR, deploy or delegate.Local changes with evidence, or a blocker.
Display and calculation agree; plans/mockups are not called completed work.
Deliver a reviewable result and clean temporary processes
Explain changes and validation so the user can decide the next step. Stop only task-created temporary processes that are no longer needed.
Summarize files, price comparison, one-shot results, actual preview URL and unknowns. Clean unnecessary processes started for this task without stopping preexisting services. Leave changes unpushed, unmerged and undeployed.A complete handoff.
Changes and evidence are reviewable with no unnecessary background process.
Find the evidence before choosing the next action
Discount changed but amount did not
Check hardcoded amounts and configuration-driven calculations.
The cloud agent pushes by default
Keep the local path while no push is required.
A screenshot exists but checks did not run
Report visual and calculation evidence separately; list unrun checks.
The task expanded into a full refactor
Return to the allowed scope and four acceptance items.
Check every item before finishing
These page-local checkboxes reset on refresh. Checking them does not verify a real service connection.
Checked 0 / 6 items
Download the inputs separately from the reference answers
Start with input and rules without showing the reference answers. After acceptance, save the report and complete prompt set.
source-data.json↓Rules and scopeDefinitions, exceptions and delivery requirements.workflow-rules.md↓Item-by-item reference resultsCompare every input with its outcome.expected-results.csv↓Complete reference reportResults, evidence and missing information.reference-report.md↓All stage prompts6 instructions with expected outputs and acceptance checks.practice-prompts.txt↓Reference implementation briefScope, price checks and execution limits.implementation-brief.md↓Check documented capabilities and exercise results separately
Content updated: · Editing a page does not reverify its sources or live workflow.
Documentation review: Unverified · no review date recorded
No case-specific documentation review is recorded. References below are provided for review and do not establish verification.
Live workflow verification: Unverified
Live accounts, external writes, scheduled execution and long-term operation are unverified. Synthetic exercises and page checkboxes do not establish live verification.
Repository snippets are synthetic. No cloud agent, real code change, PR, push or deployment was performed. The exercise stops at a local task brief.
- Cursor: Cloud Agents API ↗
Check starting state, branches, PR creation and push behavior. No cloud agent was launched here.
- Grok Bot: apps and the shared computer ↗
Check connected accounts, available tools, visibility and user takeover.
- Grok Bot: Skills and Routines ↗
Validate a manual run before saving a method or scheduling it.
Try another small batch of your own material
Retain original input and operation records. Verify that each result can be traced before adding connections or recurring runs.