Build a Reusable Lesson-Planning Skill
Create a Skill named “Pembantu RPH” that produces lesson plans in your school's format, so the workflow can be reused without repeating the same long instructions
Plan automated workflows, connect tools, and define where agents need human review
Create a Skill named “Pembantu RPH” that produces lesson plans in your school's format, so the workflow can be reused without repeating the same long instructions
Identify duplicate files and files larger than 100MB, then list their names and sizes from largest to smallest without deleting anything
Identify files that may contain personal IDs, passports, tax documents, bank statements, or medical records, and list their current sharing permissions without changing them
Find files with vague or generic names, identify their contents, and propose descriptive filenames for review before renaming
Use an approved plan to create folders and move files into their intended locations, leaving uncertain files untouched
Ask Claude to review a Drive folder and propose a clearer structure based on its files. The plan flags duplicates, unclear filenames, and misplaced files before any changes are made
Create a client-readable specification for automating a process, including its goal, trigger, inputs, logic, result, test scenarios, and success metric.
Propose no-code and coded implementation options for a process, specifying triggers, steps, data storage, what people see, and maintenance complexity.
Decide which process steps an agent can handle, which should stay with a person, and where human approval is needed; identify likely failure situations and safeguards.
Define escalation triggers and the agent’s handoff message for complaints, discount requests, repeated questions, legal references, and high-value requests.
Generate a ready-to-use system prompt covering the agent’s role, tone, prohibitions, response format, and handling of missing information.
Create a concrete table of what a business agent handles, what it must stop on, and when it hands a request to a person.
Create a plain-language operating guide that explains what a workflow owner sees, handles manually, checks when nothing arrives, and whom to contact.
Check a scenario for common failure cases and specify what it does and whom it notifies in each case.
Specify which fields move between two steps, their formats, and the key used to connect them, using only existing tools.
Compare requested automations with the cost and complexity of handling rare cases, exceptions, and decisions requiring human judgment.
Break one process step into an ordered automation scenario, including its trigger, actions, conditions, and missing-data handling.
Classify each process step by how suitable it is for automation, using only tools already in place.
Manage a versioned prompt vault with strict file contracts, Git-based versioning, an INDEX.md catalog, and defined operating modes.
A prompt-generation agent that applies multiple thinking styles and structured prompt-design guidance to create tailored, effective prompts.
Investigate factuality failures across context, retrieval, prompts, tools, and workflow design. Identify the likely root cause, recommend a targeted fix, and define verification cases.
Review 50 agent conversations, identify five incorrect responses, group them by cause, and propose concrete system-prompt rules.
Create 20 challenging messages for an agent and specify what it should say instead of inventing information.
Create a ready-to-use system prompt covering an agent's role, boundaries, company knowledge, tone, response format, prohibited topics, and handoff language.
Convert a one-sentence task into an ordered technical specification, identify ambiguities, and ask clarifying questions.
Review 50 conversations, identify five incorrect responses, group them by cause, and suggest specific system-prompt rules to fix them.
Create a hand-fillable checklist for testing an agent against 10 typical scenarios and 5 edge cases.
Generate 20 challenging messages and specify how the agent should respond without inventing information.
Convert a task request into an ordered technical brief, highlighting ambiguities and questions for the requester.
Identify repeatable weekly processes and decide which to automate fully, automate partially, or leave alone.
Generate ten stress tests for an agent system prompt and recommend a specific prompt revision for each expected failure.
Draft an agent system prompt that specifies its context, output format, prohibitions, and behavior when information is missing.
Create a code-free technical specification for an agent, including its inputs, logic, error handling, acceptance criteria, and test cases.
Assess routine tasks by frequency, time spent, and human-error risk, then recommend what to automate and where to begin.
Understand a scenario error, receive two hints before the fix, and learn one habit to prevent similar problems.
Turn a manual task into a modular automation plan, with testable steps, module inputs, and a weekly run guide.
Get a numbered setup checklist for connecting an automation platform to selected services, including checks and common fixes.
Identify repetitive work, estimate its monthly time cost, and prioritize tasks for automation.
Generate ten stress tests for an agent prompt, identify expected failures, and recommend specific prompt fixes.
Create an agent system prompt from a target objective, examples, and operational requirements.