Status: draft. From the Sprint 4 Experience Brief (Toolbox and Daily Briefing), the Pilot Product: Functional Scope and User Journey (section 3, The Toolbox) and what the prototype does today. Lines marked Proposed are Claude's, for Kate to confirm. Each builder has its own brief with the same seven sections; this page holds what all four share.
1. The four builders
The tools differ in what they ask and what they make, never in how they work. She learns the pattern once.
| Builder | Makes | Marked used as | Checks back | Copy and export |
|---|---|---|---|---|
| Positioning Builder | Positioning statement, bios in three lengths, leadership narrative, an introduction for a moment | Used | 7 days after | Copy each output; PDF and Word for the bios |
| Pitch Builder | Pitch, subject lines, short and expanded versions, angle and why now, follow-up; where it goes (outlet, what they ask for, contact, how to send) and four steps to send it | Sent | 5 days after | Copy to send; PDF and Word |
| Situation Brief | One-page brief, opening statement, follow-up email | Used | The day after the moment, or 2 days after | Copy; PDF and Word, one page |
| Thought Leadership Builder | LinkedIn post and hooks, or a short or long-form article with headlines | Published | 7 days after | Copy to post on LinkedIn; PDF and Word for articles |
The check-back days are the prototype's defaults. Only the Situation Brief's comes from the brief (the Maya journey's two-day follow-up after a manager check-in); the rest are Proposed.
2. One workflow
2.1 The sequence
Every builder runs the same six steps, in this order:
- Use case. One sentence on what it makes, one on when to use it, what she will need, and how it connects to her plan. No feature tour. Every concept opens on it, Interview included (since 13 Oct).
- Inputs. Only what this artifact needs.
- First draft. Editable, never presented as finished or as fact.
- Revise. Structured options and a conversational follow-up.
- Copy or export.
- Mark ready, then used. The Loop follows up from there.
2.2 Where she starts, and what arrives filled in
| Entry point | What arrives filled in |
|---|---|
| An action step (Plan, Home, the advisor) | The step, its plan area, its why this, why now and why you, and anything the step already says (a moment's type, who will be there) |
| A Briefing read (Thought Leadership Builder only) | The article's title, publisher and reason line, labelled as the source |
| The Toolbox or the library | Her direction and plan only |
| "What's on your mind?" (Toolbox Concept 1) | What she said, as the moment or the topic |
An artifact already in her library opens straight into its draft, never through the Toolbox page.
2.3 What every builder must do
- Open from every one of its entry points.
- Create the Loop record the moment the first draft exists, not before.
- Keep the draft where she left it if she leaves part way, so the library finds it exactly as it was.
- Never present a draft as finished or as fact.
- Keep what ExecHQ wrote distinguishable from what she supplied (section 4).
3. What goes into a draft
- Ask only for what this artifact needs. Anything ExecHQ already knows is shown filled in and editable, never asked again.
- The first question sits on the use case screen, beside the start button, so the steps after it don't repeat it. When a plan step has already answered it, it arrives chosen and says so.
- Required and optional. A builder asks the questions only she can answer, one at a time in the guided concepts. What ExecHQ already has, and anything optional, waits on one last screen to check.
- Enrichers are offers, never gates. When something would improve the draft ("Add three recent accomplishments"), it is offered inside the flow with why, and she can skip it. The draft never waits for one.
- She supplies the facts. In V1 nothing is imported. No LinkedIn, no company data, no scraping. Every fact in a draft is one she typed or chose.
- Private stays private. Nothing she said about her company, her manager or internal politics goes into a draft unless she puts it there herself.
4. The three kinds of wording
Every draft is made of three kinds of text, and she can always tell them apart:
| Kind | What it is | How it looks |
|---|---|---|
| Your words | What she typed or chose, placed as she said it | Primary ink |
| ExecHQ's wording | Framing around her words: openings, transitions, the structure of a section | Secondary ink, dotted underline, a key above the draft |
| A question only you can answer | A gap ExecHQ will not guess at: a result, a name, a date | A marked question in place; listed again under "Questions for you" |
Rules:
- Never invent specifics. No names, numbers, results, dates, quotes or employers she did not give. A gap is a question to her ("What result did the team deliver?"), never a plausible filler.
- Her words fit the sentence. Placed mid-sentence, her words take a small first letter (unless they start with "I" or a proper name) and lose their own full stop. ExecHQ phrases around what she said rather than forcing it into a lead-in it doesn't fit.
- Call it a draft until she marks it ready. "Here is a first draft of your brief", never "Here is your brief".
- An edited section becomes hers. Once she rewrites a section, all of it is her words.
- Answering a question puts her answer in its place, in her words.
5. Changing a draft
- Structured options: voice (plainer, warmer, more direct), specificity (broader, more specific), length (shorter, longer), audience (peers, senior leaders, my team) and emphasis. Each builder names its own emphasis choices. Pitch Builder calls voice "tone" and has no audience option, as its PRD section names them. In the prototype every builder now acts on length and emphasis, and Thought Leadership on voice and audience too; the rest are recorded with the version (no model).
- A conversational follow-up: "Or say what to change", in her words.
- Versions are saved automatically, one for every revision, with what changed.
- She can view any earlier version and restore it. Restoring makes a new version from the old one; nothing is overwritten.
- Proposed: a revision never touches a section she has rewritten herself.
6. Using it, and the Loop
6.1 Statuses
The Sprint 2 Loop statuses, identical everywhere an artifact appears (the Toolbox, the library, the Plan, Home):
| Status | Means |
|---|---|
| Draft | Created and saved; she may still be editing ("Still editing") |
| Ready to use | She says it is ready |
| Used / Sent / Published | She used it, by the kind's own word (section 1) |
| Waiting to hear | Used, and ExecHQ will check back |
| Outcome logged | She said what came of it |
| Closed | No further action needed. Never "Done" or "Complete", which belong to action steps. |
6.2 Marking it
- Mark ready, then Mark used / sent / published. Posting or sending is always something she does outside ExecHQ. Nothing implies ExecHQ posts or sends ("Copy to post on LinkedIn", never "Post to LinkedIn").
- Check back. Once used, ExecHQ asks what came of it after the builder's check-back days (section 1), unless she chose otherwise when she marked it ready.
- The question: "You sent your pitch for the Q1 planning review on Thursday. What came of it?" Five answers, one tap each: It went well, It was mixed, Not the way I hoped, Nothing yet, It's no longer relevant. A builder can word its own: the Situation Brief asks "How did it go?" (with "It hasn't happened yet" for a meeting that moved) and Thought Leadership "How did it land?", with answers to match (13 Oct). Her tap is all the feedback: ExecHQ never asks a second question, and gives the next step whatever she answers (13 Oct). Each builder's own table is under "After she answers" in its page. The outcome each answer records is the same.
- She can answer early, in the builder. Once it is used, the draft itself asks "What came of it?" with the same five answers, so she never has to wait for the check-back (the PRD's "Mark used and report outcome"). It is the Loop's own check-in, the one Home uses; answering here is the same as answering there.
- "Nothing yet" is normal, never a failure. It asks again 3 days later, then 7 days after that; after a third "Nothing yet" it stops asking, and she can still report an outcome whenever she likes. It never asks about more than one thing at a time.
- After an outcome, the next step. Whatever she answers, ExecHQ says it back and offers what to do next: one leads, with at most one more, each with one action (a Plan step to add, part of the work to copy, a new one starting from this one, another tool starting from this work, today's Briefing, a sharper version, or what's next on her Plan). A Plan step goes on her Plan only when she says so. Every builder, every concept, and the same next step wherever she answers: in the builder, on Home and to the advisor (one set of rules,
lib/next-after.ts). - A note is hers to add. Every feedback card offers "Add a note" (optional, only she sees it), alongside the next step. ExecHQ never asks for one.
- Closing. She can close anything, any time. "It's no longer relevant" closes it, except her positioning statement, which every tool draws on: there it only stops the asking.
6.3 The Signal Picture
Drafting, editing or readying something is not a signal. Only what she used, sent or published, and what came of it, appears on her Signal Picture, under the plan area the artifact serves.
7. The library and reuse
- Everything she made, in one place, most recent first, with title, tool, status, last change, versions and where it came from.
- Reuse makes a new artifact, with its own Loop record, linked back to the one it started from ("Started from your pitch for the Q1 planning review").
- Delete asks once, and says what else goes: its Loop record and any Signal Picture items it produced.
- Never empty after onboarding. Her one-sentence positioning statement is her first artifact. Only someone who left onboarding before saving sees an empty library, pointing to her first action step.
8. How the agent behaves, in every builder
Proposed, all of this section. Drawn from decisions already made in the prototype and the brand document. For Kate to confirm.
- A trusted advisor, not a writer for hire. Calm authority, short sentences, her words back to her. No "great question", no corporate hedging, no exclamation marks.
- Asks rather than guesses. When a draft needs a fact she hasn't given, it leaves a question in place. A draft with three honest questions beats one with three invented details.
- What it is for, not what it will achieve. A recommended next use says what the artifact is for ("Use it to open your 1:1"), never what it will win.
- Hers to change. It never argues with an edit, never restores its own wording over hers, and never revises a section she rewrote.
- No publishing, no sending. It prepares; she acts.
- Provenance stays. A draft that started from a Briefing read or a plan step says so, for the life of the artifact.
9. Guardrails
From the PRD (Kate, 13 October). They hold in every builder and in all three builder concepts (Guided steps, Document workspace, Interview). Each line says how the prototype keeps it.
| Guardrail | How every builder keeps it |
|---|---|
| Never publish automatically. She authorizes and performs any external posting. | Nothing in ExecHQ posts, sends or schedules. Copy is worded for what she will do herself ("Copy to send", "Copy to post on LinkedIn"). "Mark sent" and "Mark published" record what she did; they do nothing outside ExecHQ. A pitch and a post say so under their buttons: "ExecHQ never posts or sends anything for you." |
| Clearly distinguish user-provided evidence from generated framing. | Every draft marks three kinds of text (section 4): her words, ExecHQ's wording (quieter ink, dotted underline) and questions for her, with a key above the draft. The same marks hold wherever a draft appears: the full draft, Concept 3's draft as it fills in, and its collapsed card. Her evidence is always her words; ExecHQ never writes evidence, only questions asking for it. A section she rewrites becomes all hers. |
| Do not infer or export private Signal Background information without explicit, scoped permission. | No builder reads her Signal Background. Everything a draft uses is shown before the draft, with where it came from (her plan, her direction, onboarding), and can be changed. What she adds in a builder (accomplishments, risks, a pasted bio) is used in that draft only. If a later version draws on Signal Background, it asks first, names exactly what it would use, and the draft labels it. |
| Make provenance visible when a draft derives from LinkedIn-imported context or a Daily Intelligence Briefing item. | A draft started from a Briefing read shows its source (title, publisher, the passage) on the draft and in the library, for the life of the artifact ("Started from today's Briefing"). Her statement from onboarding and the plan step a draft serves ("For: …") show on the draft too. V1 imports nothing from LinkedIn (manual input only); if it ever does, the draft says "Uses your LinkedIn context" in the same place. In Concept 3 on a phone, the source stays above the open draft. |
10. Open questions
- Check-back days. Only the Situation Brief's comes from the brief. Are 7 days (positioning, posts) and 5 days (pitches) right?
- Export formats. Copy for everything; PDF and Word for the Situation Brief, the bios, pitches (Kate, 13 Oct) and articles. Posts are copy-only. Confirm.
- The follow-up question. The Situation Brief and Thought Leadership now have their own. Should Pitch ("Did you hear back?") and Positioning too?
- Emphasis choices. Each builder now has its own (13 Oct): Positioning What I lead · What I'm known for · Where I'm headed; Pitch The angle · The proof · Why now; Situation Brief The ask · The evidence · The risks; Thought Leadership My example · The article's point · The question at the end. In the prototype all four builders now act on their emphasis choices.