# The “Actually Do the Thing” Prompt Kit

Six prompts for turning a request into actual files, checked results, and an honest handoff.

**For:** Anyone whose AI keeps describing the work instead of producing the requested result.

**You will make:** A project brief, a workable small build, a verified output, and a record of what remains.

## Why use it

A useful assistant needs a target, inputs, access, and a way to recognize a finished result. “Make me something great” leaves a surprising amount of room for interpretive dance.

## What you need

- An assistant you can use; file creation is helpful but not required for planning.
- One small deliverable and any source files you are authorized to use.
- A destination for the result and any spending or publishing boundaries.

## First time? Start here

1. **Open and start:** Open an AI chat you already use, such as ChatGPT, Claude, or Gemini, and start a new conversation. Choose one small thing you want to make, such as a workshop checklist. You do not need a project brief yet.
2. **Next step:** Go to “1. Define the deliverable.” If you cannot fill its fields, copy its discovery prompt, paste it in the chat message box, and press Send or Enter. Answer one question at a time. Review the finished brief before moving to the next numbered prompt in the same conversation.
3. **Save your work:** Keep one document named workshop-project-notes, or a name for your own project. Copy the accepted brief, actual output, and final handoff into it. If a download is offered, download and open it; a chat message saying “saved” is not a file on your computer.

**Done when:** You can open the requested result and compare it with the brief’s checks. The handoff says which checks really ran and what still needs your decision.

Use these prompts as a short sequence, not six incantations pasted simultaneously. Start with the brief, then check capabilities before asking for implementation.

A prompt cannot supply an unavailable tool. If your assistant cannot create or save the requested file, it should produce a usable supported format and explain the remaining step.

## The steps

### 1. Choose one inspectable result

Name a file, page, table, or other result you could actually open. Supply its intended reader and the task it should help them complete. Use Prompt 1 to turn that into acceptance checks: concrete things you can look for in the result.

**Expected output:** A one-page project brief.

**Check:** You can say what you will inspect, rather than only how you want to feel about it.

### 2. Check the assistant’s working area

Attach the actual source material, then use Prompt 2. A link is not proof that the assistant can read it; ask it to identify the files and capabilities it actually has.

**Expected output:** A capability and input inventory.

**Check:** Missing access is visible before implementation begins.

### 3. Build the smallest useful version

Use Prompt 3 after essential inputs are present. Give the assistant permission for ordinary reversible preparation within your request and specify any actions that still need a decision.

**Expected output:** The actual draft deliverable, or a clearly identified supported partial output.

**Check:** The assistant links to or shows the output it created.

### 4. Inspect, fix, and hand off

Use Prompt 4 on the actual output, Prompt 5 for concrete defects, and Prompt 6 when stopping. Keep the files and quality-check notes together. If you have only chat text, paste the actual draft into the next prompt rather than referring to an unavailable file.

**Expected output:** A checked version and honest handoff.

**Check:** No test is marked passed merely because the assistant wrote a checklist.

## Help with the details

For the prompt you need, copy its discovery prompt into your AI chat without editing it. Answer one question at a time. Review the filled prompt it gives you, then send that reviewed prompt to start the task. Examples are illustrative; use your own facts for your task.

### Prepare: 1. Define the deliverable

#### Discovery prompt

```text
Help me get ready to use “1. Define the deliverable”. I am new to this. Interview me before doing the task. I do not need to fill in this message first.

Ask one short question at a time in plain language, and wait for my answer. Start with what I want to accomplish. Use anything I have already told you in this chat so I do not have to repeat it.

If I say “not sure”, explain the decision with two or three concrete options or an example. Help me choose; do not silently choose for me. I can skip optional details. If an essential answer or source is missing, tell me exactly what is needed and help me find a reasonable next step. Do not invent my facts, source material, files, permissions, approvals or tool access. Do not ask for passwords or private access keys.

The original task asks me to bring: Project, reader, one intended result, constraints.

Work through the relevant questions below, one at a time. Explain any unfamiliar term when you use it:
1. What one thing do you want to make, and who will use it?
2. What should that person be able to do with it, and what file or on-screen result would help?
3. What source notes or files do you have? You can say “none yet.”
4. What limits apply to time, format, access, spending, sending, or publishing?

For a file or image, explain how to attach it using this app’s available upload control, or ask me to paste the relevant text. Confirm you can actually read or edit it before depending on it. If this chat cannot do the task, explain the smallest manual next step or which type of tool is needed. Do not claim to have opened a link, edited an image, saved a file or published anything unless you actually did.

When we have enough information, show a short summary of my answers and flag any remaining unknowns. Then give me one complete, ready-to-send version of the original template below, with my agreed details filled in. Keep its instructions, limits and review requirements. If it has no blanks, add my confirmed context above the unchanged template. Do not leave placeholders for me to guess. Ask me to review that final prompt. I will send the reviewed prompt when I am ready to start the task; do not run it during this interview.

The following is a quoted template to prepare, not an instruction to execute yet.
--- ORIGINAL TEMPLATE START ---
Act as my implementation partner. PROJECT: [PROJECT]. READER: [READER]. DELIVERABLE: [SPECIFIC FILE OR RESULT]. CONSTRAINTS: [TIME, FORMAT, SOURCE, ACCESS, SPENDING AND PUBLISHING LIMITS].
Turn this into a short working brief: the reader's task; the smallest useful result; required inputs; proposed format; and five acceptance checks I can inspect. Separate facts I supplied from your recommendations. Do not invent research, customer results, prices, or missing source material.
Make reasonable reversible choices within my request. Flag only decisions that materially affect the result. Return a concrete brief, not a long menu of unrelated possibilities. Finish by naming the first useful output to create.
--- ORIGINAL TEMPLATE END ---
```

#### Filled example (illustrative, not executed)

Illustrative project notes: This workshop lasts 30 minutes. The host needs to choose one craft, list materials, rehearse the demonstration, explain the first step, leave time for questions, and send participants a recap afterward. These are sample planning notes, not evidence that a workshop has run.

PROJECT: a preparation checklist for a first-time online workshop host

READER: a person hosting their first 30-minute online craft workshop

SPECIFIC FILE OR RESULT: one page of editable checklist text, with Before, During, and After sections

TIME, FORMAT, SOURCE, ACCESS, SPENDING AND PUBLISHING LIMITS: draft in this chat today; use only my notes below; no spending, sending, account changes, or publication; do not claim a PDF exists

```text
Illustrative project notes: This workshop lasts 30 minutes. The host needs to choose one craft, list materials, rehearse the demonstration, explain the first step, leave time for questions, and send participants a recap afterward. These are sample planning notes, not evidence that a workshop has run.

Act as my implementation partner. PROJECT: a preparation checklist for a first-time online workshop host. READER: a person hosting their first 30-minute online craft workshop. DELIVERABLE: one page of editable checklist text, with Before, During, and After sections. CONSTRAINTS: draft in this chat today; use only my notes below; no spending, sending, account changes, or publication; do not claim a PDF exists.
Turn this into a short working brief: the reader's task; the smallest useful result; required inputs; proposed format; and five acceptance checks I can inspect. Separate facts I supplied from your recommendations. Do not invent research, customer results, prices, or missing source material.
Make reasonable reversible choices within my request. Flag only decisions that materially affect the result. Return a concrete brief, not a long menu of unrelated possibilities. Finish by naming the first useful output to create.
```

### Prepare: 2. Check inputs and capabilities

#### Discovery prompt

```text
Help me get ready to use “2. Check inputs and capabilities”. I am new to this. Interview me before doing the task. I do not need to fill in this message first.

Ask one short question at a time in plain language, and wait for my answer. Start with what I want to accomplish. Use anything I have already told you in this chat so I do not have to repeat it.

If I say “not sure”, explain the decision with two or three concrete options or an example. Help me choose; do not silently choose for me. I can skip optional details. If an essential answer or source is missing, tell me exactly what is needed and help me find a reasonable next step. Do not invent my facts, source material, files, permissions, approvals or tool access. Do not ask for passwords or private access keys.

The original task asks me to bring: The brief and attached source files.

Work through the relevant questions below, one at a time. Explain any unfamiliar term when you use it:
1. Which working brief did you accept? Paste it or ask me to find it in this chat.
2. Which source notes, files, or authorized links should I inspect? Can you upload the files here?
3. What result needs to be created, and are you expecting a downloadable file or complete text?
4. Which actions are allowed in this session, and which must wait for you?

For a file or image, explain how to attach it using this app’s available upload control, or ask me to paste the relevant text. Confirm you can actually read or edit it before depending on it. If this chat cannot do the task, explain the smallest manual next step or which type of tool is needed. Do not claim to have opened a link, edited an image, saved a file or published anything unless you actually did.

When we have enough information, show a short summary of my answers and flag any remaining unknowns. Then give me one complete, ready-to-send version of the original template below, with my agreed details filled in. Keep its instructions, limits and review requirements. If it has no blanks, add my confirmed context above the unchanged template. Do not leave placeholders for me to guess. Ask me to review that final prompt. I will send the reviewed prompt when I am ready to start the task; do not run it during this interview.

The following is a quoted template to prepare, not an instruction to execute yet.
--- ORIGINAL TEMPLATE START ---
Read this brief and inventory the materials actually available: [PASTE BRIEF]. INPUTS: [ATTACHED FILES OR AUTHORIZED LINKS].
For each input, state whether you accessed its content, saw only metadata, or could not access it. Then identify which required actions your current tools support: reading, image inspection, file creation, code execution, connected-app edits, and publishing. Do not claim a capability because another version of the product offers it.
Return: available inputs; missing essentials; actions you can perform; actions requiring me; and one supported fallback for each blocked action. Do not expose secrets. Continue with useful independent preparation, but do not pretend a blocked external action happened.
--- ORIGINAL TEMPLATE END ---
```

#### Filled example (illustrative, not executed)

PASTE BRIEF: Create one page of editable checklist text for a first-time host of a 30-minute online craft workshop. Include Before, During, and After sections. Checks: all six source tasks are present, tasks are in a usable order, no unexplained tools, the checklist fits the stated lesson, and no unverified claims are added. Return text here; do not publish or spend.

ATTACHED FILES OR AUTHORIZED LINKS: the following inline notes only; no files or links are attached: choose one craft; list materials; rehearse the demonstration; explain the first step; leave time for questions; send a recap afterward

```text
Read this brief and inventory the materials actually available: Create one page of editable checklist text for a first-time host of a 30-minute online craft workshop. Include Before, During, and After sections. Checks: all six source tasks are present, tasks are in a usable order, no unexplained tools, the checklist fits the stated lesson, and no unverified claims are added. Return text here; do not publish or spend.. INPUTS: the following inline notes only; no files or links are attached: choose one craft; list materials; rehearse the demonstration; explain the first step; leave time for questions; send a recap afterward.
For each input, state whether you accessed its content, saw only metadata, or could not access it. Then identify which required actions your current tools support: reading, image inspection, file creation, code execution, connected-app edits, and publishing. Do not claim a capability because another version of the product offers it.
Return: available inputs; missing essentials; actions you can perform; actions requiring me; and one supported fallback for each blocked action. Do not expose secrets. Continue with useful independent preparation, but do not pretend a blocked external action happened.
```

### Prepare: 3. Make the small version

#### Discovery prompt

```text
Help me get ready to use “3. Make the small version”. I am new to this. Interview me before doing the task. I do not need to fill in this message first.

Ask one short question at a time in plain language, and wait for my answer. Start with what I want to accomplish. Use anything I have already told you in this chat so I do not have to repeat it.

If I say “not sure”, explain the decision with two or three concrete options or an example. Help me choose; do not silently choose for me. I can skip optional details. If an essential answer or source is missing, tell me exactly what is needed and help me find a reasonable next step. Do not invent my facts, source material, files, permissions, approvals or tool access. Do not ask for passwords or private access keys.

The original task asks me to bring: Accepted brief, available inputs, output destination.

Work through the relevant questions below, one at a time. Explain any unfamiliar term when you use it:
1. Which brief and acceptance checks should I follow?
2. What source information is actually available in this chat or through uploaded files?
3. What exact output do you want now, and where should it be saved or displayed?
4. Is anything still blocked by missing access, cost, or a decision?

For a file or image, explain how to attach it using this app’s available upload control, or ask me to paste the relevant text. Confirm you can actually read or edit it before depending on it. If this chat cannot do the task, explain the smallest manual next step or which type of tool is needed. Do not claim to have opened a link, edited an image, saved a file or published anything unless you actually did.

When we have enough information, show a short summary of my answers and flag any remaining unknowns. Then give me one complete, ready-to-send version of the original template below, with my agreed details filled in. Keep its instructions, limits and review requirements. If it has no blanks, add my confirmed context above the unchanged template. Do not leave placeholders for me to guess. Ask me to review that final prompt. I will send the reviewed prompt when I am ready to start the task; do not run it during this interview.

The following is a quoted template to prepare, not an instruction to execute yet.
--- ORIGINAL TEMPLATE START ---
Use the agreed brief and accessible source files to create [OUTPUT]. Start with the smallest version that satisfies its acceptance checks. Follow any documented source workflow in order, adapting only where necessary and labeling your additions.
Do the authorized work with your available tools. Save actual deliverables where supported and show their locations. If file creation is unavailable, provide the complete usable content in a supported form and label that limitation. Do not substitute a description of what you would create for the creation itself.
Keep source material intact. Do not spend money, publish, send messages, or change access unless that action is already authorized. Stop only for essential missing inputs or decisions that require my judgment. Report completed work, outputs, and open issues.
--- ORIGINAL TEMPLATE END ---
```

#### Filled example (illustrative, not executed)

Illustrative accepted brief: Help a first-time host prepare a 30-minute online craft workshop. Source notes: choose one craft; list materials; rehearse the demonstration; explain the first step; leave time for questions; send a recap afterward. Acceptance checks: cover all six tasks, use the three named sections, keep the sequence usable, explain any unfamiliar term, and add no unverified claims. I am requesting a text draft only. There are no attached files and no permission to publish, send, or spend.

OUTPUT: one page of editable workshop-preparation checklist text in this chat, with Before, During, and After sections

```text
Illustrative accepted brief: Help a first-time host prepare a 30-minute online craft workshop. Source notes: choose one craft; list materials; rehearse the demonstration; explain the first step; leave time for questions; send a recap afterward. Acceptance checks: cover all six tasks, use the three named sections, keep the sequence usable, explain any unfamiliar term, and add no unverified claims. I am requesting a text draft only. There are no attached files and no permission to publish, send, or spend.

Use the agreed brief and accessible source files to create one page of editable workshop-preparation checklist text in this chat, with Before, During, and After sections. Start with the smallest version that satisfies its acceptance checks. Follow any documented source workflow in order, adapting only where necessary and labeling your additions.
Do the authorized work with your available tools. Save actual deliverables where supported and show their locations. If file creation is unavailable, provide the complete usable content in a supported form and label that limitation. Do not substitute a description of what you would create for the creation itself.
Keep source material intact. Do not spend money, publish, send messages, or change access unless that action is already authorized. Stop only for essential missing inputs or decisions that require my judgment. Report completed work, outputs, and open issues.
```

### Prepare: 4. Check the actual result

#### Discovery prompt

```text
Help me get ready to use “4. Check the actual result”. I am new to this. Interview me before doing the task. I do not need to fill in this message first.

Ask one short question at a time in plain language, and wait for my answer. Start with what I want to accomplish. Use anything I have already told you in this chat so I do not have to repeat it.

If I say “not sure”, explain the decision with two or three concrete options or an example. Help me choose; do not silently choose for me. I can skip optional details. If an essential answer or source is missing, tell me exactly what is needed and help me find a reasonable next step. Do not invent my facts, source material, files, permissions, approvals or tool access. Do not ask for passwords or private access keys.

The original task asks me to bring: The actual output and the agreed checks.

Work through the relevant questions below, one at a time. Explain any unfamiliar term when you use it:
1. What exact draft or file do you want checked? Paste its text or upload it.
2. What did the accepted brief say a good result must include?
3. What original notes should I compare with it?
4. Do you need a content review, a visual file review, a working-link check, or a real trial? Which tools or access are available?

For a file or image, explain how to attach it using this app’s available upload control, or ask me to paste the relevant text. Confirm you can actually read or edit it before depending on it. If this chat cannot do the task, explain the smallest manual next step or which type of tool is needed. Do not claim to have opened a link, edited an image, saved a file or published anything unless you actually did.

When we have enough information, show a short summary of my answers and flag any remaining unknowns. Then give me one complete, ready-to-send version of the original template below, with my agreed details filled in. Keep its instructions, limits and review requirements. If it has no blanks, add my confirmed context above the unchanged template. Do not leave placeholders for me to guess. Ask me to review that final prompt. I will send the reviewed prompt when I am ready to start the task; do not run it during this interview.

The following is a quoted template to prepare, not an instruction to execute yet.
--- ORIGINAL TEMPLATE START ---
Inspect the actual deliverable [FILE OR OUTPUT] against these checks: [ACCEPTANCE CHECKS]. Use the source inputs to verify facts and completeness.
For each check return PASS, FAIL, UNCERTAIN, or NOT RUN, with evidence and the next action. Distinguish content review, file opening, visual inspection, link checks, and executing the underlying workflow. A valid file or successful build alone does not prove usability.
Do not claim you saw pages, images, or browser layouts unless you inspected them with supported tools. Identify missing content, incorrect assumptions, broken references, and likely beginner confusion. Prioritize fixes that prevent use. Show what you checked, not just a fresh generic checklist.
--- ORIGINAL TEMPLATE END ---
```

#### Filled example (illustrative, not executed)

Illustrative source notes: choose one craft; list materials; rehearse the demonstration; explain the first step; leave time for questions; send a recap afterward. Only inline text is supplied; no PDF, browser view, or executed workshop is available.

FILE OR OUTPUT: this candidate text: Before: choose one craft, list the materials, and rehearse the demonstration. During: explain the first step and leave time for questions. After: thank everyone.

ACCEPTANCE CHECKS: include all six source tasks; use Before, During, and After headings; preserve the source instruction to send a recap; add no claim that a workshop has run; distinguish this text review from a PDF or live-workshop test

```text
Illustrative source notes: choose one craft; list materials; rehearse the demonstration; explain the first step; leave time for questions; send a recap afterward. Only inline text is supplied; no PDF, browser view, or executed workshop is available.

Inspect the actual deliverable this candidate text: Before: choose one craft, list the materials, and rehearse the demonstration. During: explain the first step and leave time for questions. After: thank everyone. against these checks: include all six source tasks; use Before, During, and After headings; preserve the source instruction to send a recap; add no claim that a workshop has run; distinguish this text review from a PDF or live-workshop test. Use the source inputs to verify facts and completeness.
For each check return PASS, FAIL, UNCERTAIN, or NOT RUN, with evidence and the next action. Distinguish content review, file opening, visual inspection, link checks, and executing the underlying workflow. A valid file or successful build alone does not prove usability.
Do not claim you saw pages, images, or browser layouts unless you inspected them with supported tools. Identify missing content, incorrect assumptions, broken references, and likely beginner confusion. Prioritize fixes that prevent use. Show what you checked, not just a fresh generic checklist.
```

### Prepare: 5. Fix only the named problems

#### Discovery prompt

```text
Help me get ready to use “5. Fix only the named problems”. I am new to this. Interview me before doing the task. I do not need to fill in this message first.

Ask one short question at a time in plain language, and wait for my answer. Start with what I want to accomplish. Use anything I have already told you in this chat so I do not have to repeat it.

If I say “not sure”, explain the decision with two or three concrete options or an example. Help me choose; do not silently choose for me. I can skip optional details. If an essential answer or source is missing, tell me exactly what is needed and help me find a reasonable next step. Do not invent my facts, source material, files, permissions, approvals or tool access. Do not ask for passwords or private access keys.

The original task asks me to bring: Current version, source of truth, specific defects.

Work through the relevant questions below, one at a time. Explain any unfamiliar term when you use it:
1. What is the current draft? Paste it or upload the exact version.
2. What original notes or accepted requirements should stay authoritative?
3. What specific problem did you observe, and where is it?
4. Which parts should stay unchanged, and what format or filename should the revision use?

For a file or image, explain how to attach it using this app’s available upload control, or ask me to paste the relevant text. Confirm you can actually read or edit it before depending on it. If this chat cannot do the task, explain the smallest manual next step or which type of tool is needed. Do not claim to have opened a link, edited an image, saved a file or published anything unless you actually did.

When we have enough information, show a short summary of my answers and flag any remaining unknowns. Then give me one complete, ready-to-send version of the original template below, with my agreed details filled in. Keep its instructions, limits and review requirements. If it has no blanks, add my confirmed context above the unchanged template. Do not leave placeholders for me to guess. Ask me to review that final prompt. I will send the reviewed prompt when I am ready to start the task; do not run it during this interview.

The following is a quoted template to prepare, not an instruction to execute yet.
--- ORIGINAL TEMPLATE START ---
Revise [CURRENT OUTPUT] using [SOURCE OF TRUTH]. Fix these observed defects: [SPECIFIC DEFECTS AND EVIDENCE]. Preserve the approved parts, tone, facts, and file format.
First identify which sections or files need to change. Make the smallest complete correction that resolves each defect. Do not invent missing evidence or replace an unknown fact with a confident guess. Save a new version or use the established versioning method, keeping the source intact.
Recheck the affected acceptance criteria. Return the revised output, what changed, checks actually performed, and any unresolved problem. If a requested change conflicts with an approved requirement, explain the conflict before making a consequential choice.
--- ORIGINAL TEMPLATE END ---
```

#### Filled example (illustrative, not executed)

CURRENT OUTPUT: the candidate checklist: Before: choose one craft, list materials, rehearse. During: explain the first step, leave time for questions. After: thank everyone.

SOURCE OF TRUTH: my planning notes: choose one craft; list materials; rehearse the demonstration; explain the first step; leave time for questions; send a recap afterward

SPECIFIC DEFECTS AND EVIDENCE: the After section omits “send a recap,” which appears in my source notes. Add that task. Keep the other five tasks and the three headings unchanged; return revision 2 as complete text here

```text
Revise the candidate checklist: Before: choose one craft, list materials, rehearse. During: explain the first step, leave time for questions. After: thank everyone. using my planning notes: choose one craft; list materials; rehearse the demonstration; explain the first step; leave time for questions; send a recap afterward. Fix these observed defects: the After section omits “send a recap,” which appears in my source notes. Add that task. Keep the other five tasks and the three headings unchanged; return revision 2 as complete text here. Preserve the approved parts, tone, facts, and file format.
First identify which sections or files need to change. Make the smallest complete correction that resolves each defect. Do not invent missing evidence or replace an unknown fact with a confident guess. Save a new version or use the established versioning method, keeping the source intact.
Recheck the affected acceptance criteria. Return the revised output, what changed, checks actually performed, and any unresolved problem. If a requested change conflicts with an approved requirement, explain the conflict before making a consequential choice.
```

### Prepare: 6. Leave a useful handoff

#### Discovery prompt

```text
Help me get ready to use “6. Leave a useful handoff”. I am new to this. Interview me before doing the task. I do not need to fill in this message first.

Ask one short question at a time in plain language, and wait for my answer. Start with what I want to accomplish. Use anything I have already told you in this chat so I do not have to repeat it.

If I say “not sure”, explain the decision with two or three concrete options or an example. Help me choose; do not silently choose for me. I can skip optional details. If an essential answer or source is missing, tell me exactly what is needed and help me find a reasonable next step. Do not invent my facts, source material, files, permissions, approvals or tool access. Do not ask for passwords or private access keys.

The original task asks me to bring: Current files, QA notes, blockers.

Work through the relevant questions below, one at a time. Explain any unfamiliar term when you use it:
1. Which project are we pausing, and what was its agreed scope?
2. Where are the actual source material and latest output, or are they only in this chat?
3. What checks were actually run and what remains untested or blocked?
4. What should happen next, and where may the continuation note be saved?

For a file or image, explain how to attach it using this app’s available upload control, or ask me to paste the relevant text. Confirm you can actually read or edit it before depending on it. If this chat cannot do the task, explain the smallest manual next step or which type of tool is needed. Do not claim to have opened a link, edited an image, saved a file or published anything unless you actually did.

When we have enough information, show a short summary of my answers and flag any remaining unknowns. Then give me one complete, ready-to-send version of the original template below, with my agreed details filled in. Keep its instructions, limits and review requirements. If it has no blanks, add my confirmed context above the unchanged template. Do not leave placeholders for me to guess. Ask me to review that final prompt. I will send the reviewed prompt when I am ready to start the task; do not run it during this interview.

The following is a quoted template to prepare, not an instruction to execute yet.
--- ORIGINAL TEMPLATE START ---
Create a continuation note for [PROJECT] based only on work actually completed. Include: goal; agreed scope; source locations; deliverable locations and versions; completed actions; checks run and results; failed or unrun checks; outstanding decisions; and the next concrete action.
Keep “written,” “source reviewed,” “tested,” and “published” as separate statuses. Do not mark my approval unless I gave it. Do not include credentials, private access tokens, or unnecessary personal information.
Make the note sufficient for another session to continue without recreating finished work. If your tools support saving it in the authorized project folder, save it and read it back. Otherwise provide the full note here and explain that it was not saved externally.
--- ORIGINAL TEMPLATE END ---
```

#### Filled example (illustrative, not executed)

Illustrative state to document, not a claim about work performed by this assistant: the source notes and candidate below are supplied inline. Source notes: choose craft, list materials, rehearse, explain first step, leave question time, send recap. Candidate: Before—choose craft, list materials, rehearse. During—explain first step, leave question time. After—send recap. No external file, source-review report, visual check, participant trial, publication, or user approval is supplied. Next action: have me inspect the wording, then manually copy it into a document. Return the continuation note in this chat; no external destination is authorized.

PROJECT: the illustrative 30-minute craft workshop checklist described below

```text
Illustrative state to document, not a claim about work performed by this assistant: the source notes and candidate below are supplied inline. Source notes: choose craft, list materials, rehearse, explain first step, leave question time, send recap. Candidate: Before—choose craft, list materials, rehearse. During—explain first step, leave question time. After—send recap. No external file, source-review report, visual check, participant trial, publication, or user approval is supplied. Next action: have me inspect the wording, then manually copy it into a document. Return the continuation note in this chat; no external destination is authorized.

Create a continuation note for the illustrative 30-minute craft workshop checklist described below based only on work actually completed. Include: goal; agreed scope; source locations; deliverable locations and versions; completed actions; checks run and results; failed or unrun checks; outstanding decisions; and the next concrete action.
Keep “written,” “source reviewed,” “tested,” and “published” as separate statuses. Do not mark my approval unless I gave it. Do not include credentials, private access tokens, or unnecessary personal information.
Make the note sufficient for another session to continue without recreating finished work. If your tools support saving it in the authorized project folder, save it and read it back. Otherwise provide the full note here and explain that it was not saved externally.
```

## Prompt toolbox

### 1. Define the deliverable

Turn an idea into a clear implementation brief.

Inputs: Project, reader, one intended result, constraints.

Example: [PROJECT] = a one-page PDF checklist for women preparing their first workshop.

Expected output: Brief with acceptance checks and unresolved decisions.

```text
Act as my implementation partner. PROJECT: [PROJECT]. READER: [READER]. DELIVERABLE: [SPECIFIC FILE OR RESULT]. CONSTRAINTS: [TIME, FORMAT, SOURCE, ACCESS, SPENDING AND PUBLISHING LIMITS].
Turn this into a short working brief: the reader's task; the smallest useful result; required inputs; proposed format; and five acceptance checks I can inspect. Separate facts I supplied from your recommendations. Do not invent research, customer results, prices, or missing source material.
Make reasonable reversible choices within my request. Flag only decisions that materially affect the result. Return a concrete brief, not a long menu of unrelated possibilities. Finish by naming the first useful output to create.
```

### 2. Check inputs and capabilities

Find out what the assistant can actually do here.

Inputs: The brief and attached source files.

Example: [INPUTS] = workshop outline.docx and logo.png, attached in this chat.

Expected output: File inventory, available actions, blockers, and workarounds.

```text
Read this brief and inventory the materials actually available: [PASTE BRIEF]. INPUTS: [ATTACHED FILES OR AUTHORIZED LINKS].
For each input, state whether you accessed its content, saw only metadata, or could not access it. Then identify which required actions your current tools support: reading, image inspection, file creation, code execution, connected-app edits, and publishing. Do not claim a capability because another version of the product offers it.
Return: available inputs; missing essentials; actions you can perform; actions requiring me; and one supported fallback for each blocked action. Do not expose secrets. Continue with useful independent preparation, but do not pretend a blocked external action happened.
```

### 3. Make the small version

Produce an actual output from the agreed brief.

Inputs: Accepted brief, available inputs, output destination.

Example: [OUTPUT] = an editable Markdown checklist plus a plain-text version.

Expected output: Actual files or fully rendered supported output.

```text
Use the agreed brief and accessible source files to create [OUTPUT]. Start with the smallest version that satisfies its acceptance checks. Follow any documented source workflow in order, adapting only where necessary and labeling your additions.
Do the authorized work with your available tools. Save actual deliverables where supported and show their locations. If file creation is unavailable, provide the complete usable content in a supported form and label that limitation. Do not substitute a description of what you would create for the creation itself.
Keep source material intact. Do not spend money, publish, send messages, or change access unless that action is already authorized. Stop only for essential missing inputs or decisions that require my judgment. Report completed work, outputs, and open issues.
```

### 4. Check the actual result

Compare the saved deliverable with its acceptance criteria.

Inputs: The actual output and the agreed checks.

Example: [CHECKS] = all seven steps present, readable headings, correct contact link.

Expected output: Evidence-based pass/fail/uncertain report.

```text
Inspect the actual deliverable [FILE OR OUTPUT] against these checks: [ACCEPTANCE CHECKS]. Use the source inputs to verify facts and completeness.
For each check return PASS, FAIL, UNCERTAIN, or NOT RUN, with evidence and the next action. Distinguish content review, file opening, visual inspection, link checks, and executing the underlying workflow. A valid file or successful build alone does not prove usability.
Do not claim you saw pages, images, or browser layouts unless you inspected them with supported tools. Identify missing content, incorrect assumptions, broken references, and likely beginner confusion. Prioritize fixes that prevent use. Show what you checked, not just a fresh generic checklist.
```

### 5. Fix only the named problems

Correct a draft without quietly changing everything else.

Inputs: Current version, source of truth, specific defects.

Example: [DEFECT] = the checklist mentions a worksheet that is not included.

Expected output: A revised output with a focused change record.

```text
Revise [CURRENT OUTPUT] using [SOURCE OF TRUTH]. Fix these observed defects: [SPECIFIC DEFECTS AND EVIDENCE]. Preserve the approved parts, tone, facts, and file format.
First identify which sections or files need to change. Make the smallest complete correction that resolves each defect. Do not invent missing evidence or replace an unknown fact with a confident guess. Save a new version or use the established versioning method, keeping the source intact.
Recheck the affected acceptance criteria. Return the revised output, what changed, checks actually performed, and any unresolved problem. If a requested change conflicts with an approved requirement, explain the conflict before making a consequential choice.
```

### 6. Leave a useful handoff

Make the next session start from the actual state.

Inputs: Current files, QA notes, blockers.

Example: [PROJECT] = workshop checklist, ready for visual review but not published.

Expected output: A concise continuation note.

```text
Create a continuation note for [PROJECT] based only on work actually completed. Include: goal; agreed scope; source locations; deliverable locations and versions; completed actions; checks run and results; failed or unrun checks; outstanding decisions; and the next concrete action.
Keep “written,” “source reviewed,” “tested,” and “published” as separate statuses. Do not mark my approval unless I gave it. Do not include credentials, private access tokens, or unnecessary personal information.
Make the note sufficient for another session to continue without recreating finished work. If your tools support saving it in the authorized project folder, save it and read it back. Otherwise provide the full note here and explain that it was not saved externally.
```

## Worked example: Illustrative example: a workshop checklist

Illustrative example, not an executed result.

**Before:** “Can you make me something to help people plan a workshop?”

1. Define one deliverable: an editable, one-page pre-workshop checklist for a first-time host.
2. Attach the workshop outline and ask the assistant to confirm it can read it.
3. Create the checklist; inspect its steps against the outline; fix the missing materials section.

**After:** Expected result: a checklist file with a clearly marked review status, not a promise that the workshop will succeed.

**Check:** Open the actual file. Can a host identify what to prepare, in what order, and what still needs a decision?

## Troubleshooting

- **The assistant keeps planning.** Name the exact output and ask Prompt 3 to create it with the tools currently available.
- **It says it saved a file but gives no file.** Ask for its actual location and a read-back. If saving was unavailable, request the complete content and correct the status.
- **The task needs an unavailable app.** Prepare inputs and a handoff for that step; do not treat a simulated action as a real app update.

## Key takeaways

- Define the output before delegating.
- Tool access is an input, not an assumption.
- Inspect the result separately from the assistant’s completion message.

## Final checklist

- [ ] The output is named and accessible.
- [ ] Source facts are preserved.
- [ ] Actual checks and remaining limits are recorded.
- [ ] Publishing and spending stayed within my authorization.

## Sources and limits

These are original, provider-independent prompt templates based on our documented production practices. Examples are illustrative. Tool access, output quality, and available file actions depend on your assistant and account.

Editorially reviewed prompt library. No end-to-end execution of these newly written prompts or the illustrative example was performed for this library. Test the actual outputs in your environment before using them.

- [Unstuck Guide Studio production method](/free-guides/instagram-saves-to-ai-playbooks): SOURCE: existing project method. The portable prompts below are newly written RECOMMENDED ADDITIONS, not quotations from a tool provider.

## Your next step

Use Prompt 1 on one small deliverable you actually need this week.
