Lunch, learn, launch: the workflow

What it does: creates a new Project as the campaign workspace, then asks Claude to produce both a cleaned brief and Project Instructions in a single pass. This forces assumptions into the open before any tactics are drafted.

Why it matters: most briefs are incomplete. Clients assume you know their sector, their constraints, their success metrics. This surfaces all of it before you start building.

Prompt

Based on this brief, generate comprehensive Project Instructions for this
campaign workspace that will also function as a cleaned brief summary. Include:
1. CAMPAIGN BRIEF SUMMARY (1 paragraph with reasonable assumptions)
2. KEY ASSUMPTIONS MADE (what you've assumed vs what was provided)
3. DEFINITION OF DONE (what does 'finished' look like?)
4. PROJECT INSTRUCTIONS:
   - Strategic context
   - Rules for assumptions (use [BRACKETS] where the user can clarify detail)
   - UK English throughout
   - Reference back to the Foundational Comms Structure once it exists
   - Maintain tone consistency with any tone rules provided
   - Flag risks and unsupported claims proactively
   - Be decisive, do not ask clarifying questions
Make it thorough but concise.

The Definition of Done tends to fall into four natural layers: strategy, content, activation, commercial outcome. Watch for those in the output, and correct any that look thin.

The strategic spine. Every later decision checks back against this. Teams that skip the spine and jump straight to tactics end up with scattered campaigns. Save the output as a file in Project Knowledge: it becomes the bank of knowledge the rest of the workflow references.

Prompt

Create a Foundational Comms Structure in exactly this format:
OBJECTIVE:
PRIMARY AUDIENCE:
SECONDARY AUDIENCE:
AUDIENCE INSIGHT: (Why would they care?)
KEY MESSAGE: (If you had one sentence, what would you say?)
SUPPORTING POINTS: (Three reasons why, with proof or credibility.)
CALL TO ACTION:
ELEVATOR PITCH: (30 seconds, conversational)
TONE RULES: (How should this sound?)
RISKS TO AVOID:
RULES:
- Use [BRACKETS] when human input is needed
- Be specific, no generic strategy-speak
- No em-dashes
Output this as a document I can add to the Project's Knowledge.

Skills are reusable instructions that apply automatically across every conversation in the Project. Anchoring tone to a recognisable brand voice makes the rule concrete and testable. Claude even has a Skill for building Skills.

Run it as two passes. First, ask for five voice options (in the live session, the room added their own suggestions in chat alongside Claude's five). Then once you have a winner, generate the Skill itself.

Prompt A: voice references

Based on the tone rules in the Foundational Comms Structure, suggest 5 real
companies whose brand voice matches what we're aiming for. Recognisable names.
Format:
1. [Company name]: [one sentence on why their voice fits]

Prompt B: build the Skill

Now create a reusable Claude Skill for this Project, based on the winning voice.
Name the Skill: "[Voice name] tone of voice". Once added to the Project, it should
apply automatically to every text generation. Include:
- VOICE PRINCIPLES (3 to 5, concrete, not generic)
- GUARDRAILS (what to avoid)
- FORMATTING RULES (UK English, sentence case, active voice, no em-dashes)
- SAMPLE SENTENCE in this voice addressing our primary audience
Base it on the tone rules in the Foundational Comms Structure and the brand voice
of [chosen company]. Output as a ready-to-paste Skill description for the Skill builder.

Paste the output into the Skill builder in the right-hand panel of your Project, and save. From that point every draft inherits the voice automatically.

Most teams jump to "let's do a LinkedIn post" without asking why LinkedIn for this audience. This prompt forces channel choice to be strategic. Eight angles, rather than one or two, because AI is good at range, and gives you genuine options to pick, merge or discard.

Prompt

Using the Foundational Comms Structure, recommend:
1. CHANNEL MIX (primary + secondary channels)
   - Why each channel for this audience
   - Best content format per channel
   - Realistic cadence
2. EIGHT CONTENT ANGLES / HOOKS
   For each: hook line, why it works for this audience, best channel,
   and how it connects to the key message.
Assume a 3-month campaign timeline. Every angle should ladder back to the
key message in the Foundational Comms Structure.

One run produces a consistent set of assets in one voice. Pick an angle, paste the prompt, and let the Project's context plus the tone Skill do their work. AI gets a bad name for slop, usually because inputs were vague. We have defined things precisely.

Prompt

Use content angle [X] to draft the following assets, checking each against
the Foundational Comms Structure, channel plan and tone of voice:
1. LinkedIn post (120-150 words)
2. LinkedIn carousel copy (5 slides, punchy)
3. Email to prospects (200-250 words)
4. Press release (350 words)
5. Internal Slack team post (informal, motivating)
6. [Leave open: let Claude choose an appropriate final asset with some flair]
For each asset:
- Match it to the recommended channel and format
- Apply the tone rules from the Foundational Structure
- Include [BRACKETS] for anything needing confirmation
- Flag any claims that need evidence

If the assets come back with too many placeholders, run this follow-up:

Redraft all six assets with these changes:
- Make reasonable assumptions rather than using [BRACKETS]
- Be decisive about specifics (dates, numbers, examples) from the brief and context
- For asset 6, make it more creative, something that would stand out
- Keep everything aligned to the Foundational Structure and tone rules

AI critiquing its own work is a real governance layer. Two layers, in fact: strategic alignment (does it still serve the Structure?) and tone consistency (does it still sound like the Skill?). The pre-publish checklist saved at the end becomes reusable across every campaign you run in this Project.

Prompt

You're now the risk and quality checker. Review all the assets you just
drafted against the Foundational Comms Structure and tone rules. Identify:
1. CLAIMS THAT NEED EVIDENCE
2. TONE MISMATCHES
3. ANYTHING THAT COULD MISLEAD OR CREATE REPUTATIONAL RISK
4. LINES TOO VAGUE OR GENERIC TO BE USEFUL
For each issue: severity (high / medium / low), quote the problem line,
explain why it's an issue, and provide a revised version.
Then:
- Rewrite the three riskiest lines
- Create a 10-POINT PRE-PUBLISH CHECKLIST reusable for every campaign

In real work, run a second opinion through ChatGPT or Gemini as well. Different models catch different issues. You always remain the editor.

The bookend. We opened with Claude generating a deck. We close with Claude generating one more, and the Project's accumulated context makes this second pass noticeably richer. It is a small move that lands the compounding-value point hard.

Prompt

Create a 1-slide recap of what we just built, in presentation format.
TITLE: "What We Built Live"
Sections:
- Project workspace created
- Foundational structure established
- 5-step workflow followed
- Custom Skill created (tone of voice)
- Key deliverables drafted for refinement
- Pre-publish checklist saved
- What this team can do tomorrow
Keep it visual and concise.